Thesis

Transcription

Thesis
CHROMATIC RESPONSES IN HUMAN VISUAL CORTEXEffects of color, velocity and contrast
Dissertation
zur Erlangung des akademischen Grades
doctor rerum naturalium
(Dr. rer. nat.)
genehmigt durch
die Fakultät für Naturwissenschaften
der Otto-von-Guericke-Universität Magdeburg
von Dipl. Psych. Marcus Grüschow
geb. am 08. 09. 1976 in Magdeburg
Gutachter:
Prof. Dr. med. Ariel Schoenfeld
Prof. Dr. Karl R. Gegenfurtner
Eingereicht am 26.02.2007
Verteidigt am 04.06.2007
Acknowledgements
I express my deepest gratitude to my supervisor Dr. Jochem Rieger. His friendship,
knowledge and precision have been inspiring. He truly conveys what it means to be a
scientist.
I would like to thank the doctoral committee and the two final reviewers. I thank Prof.
Thomas Münte, Prof. Christoph Herrmann, PD Ariel Schönfeld, Prof. Hans-Jochen Heinze
and Prof. Herman Hinrichs for their support throughout the last years.
I extend my thanks to Prof. Bob Fendrich who has always been supportive by means
of hilarious, but even more so, ingenious comments.
A very special thank you belongs to my two colleagues Franziska Schalk and Nick
Köchy, with whom the perception group experience has truly been one. Their friendship and
encouragement has been invaluable. The dude obeys.
I am also very grateful for the stimulating social environment that seems natural to the
Department of Neurology II at the University Clinic Magdeburg. Big props go to Nico
Boehler, Laura Hermann, Jacqueline Lemme, Sascha Tyll, Daniel Bergmann, Bjoern Bonath,
Toemme Noesselt, Uwe Mattler, Yariv Festman, Oussama Hamid, Hartmut Schütze, Claus
Tempelmann, Jörn Kaufmann, Nils Bodammer, Martin Kanowski and Michael Scholz.
My work could not have been completed without the continuous support from
individuals outside of the science world. I owe to the members of my Bands Kantona and
Enrico and I thank the TRF. Thanx to Lars, Raul, Peter, Alexis and my dear Anna.
Finally, I would like to thank my parents. Their unconditional support has given me
the strength to follow my dreams and allows me to go beyond.
Contents
1 GENERAL INTRODUCTION............................................................................... 1
1.1
A short history of color science ....................................................................1
1.2
Neurophysiological basics.............................................................................4
1.2.1
Anatomy and function of the visual system ...................................................4
1.2.1.1
Three receptor classes ..............................................................................5
1.2.1.2
Neurophysiological implementation of color-opponency ............................6
1.2.2
Spatial response properties of visual neurons...............................................8
1.2.2.1
Receptive fields of retinal ganglion cells ....................................................8
1.2.2.2
Receptive fields of primary visual cortex (V1) ............................................8
1.2.2.3
Retinotopic organization ............................................................................9
1.2.3
Anatomy and function of geniculo-cortical pathways...................................10
1.2.3.1
Anatomy...................................................................................................10
1.2.3.2
Functional specialization in the LGN........................................................11
1.2.3.3
Functionally specialized pathways in the cortex ......................................12
1.2.3.4
Cortical color mechanisms.......................................................................13
1.2.3.5
Cortical color center ? ..............................................................................15
1.2.3.6
Chromatic responses in human cortex ....................................................17
1.3
Experimental objectives ...............................................................................17
2 GENERAL METHODS ...................................................................................... 20
2.1
Stimulus control............................................................................................20
2.1.1
Classifying colors ........................................................................................20
2.1.1.1
Physiological color spaces.......................................................................20
2.1.1.2
Luminance ...............................................................................................22
2.1.1.3
Isoluminance............................................................................................24
2.1.1.4
Luminance vs. achromatic stimuli ............................................................25
2.1.1.5
Color vs. luminance .................................................................................25
2.1.2
Computing cone-contrast colors..................................................................26
2.1.2.1
Light to cone excitations ..........................................................................26
2.1.2.2
Cone excitations to light...........................................................................29
2.2
Magnetic resonance imaging.......................................................................31
2.2.1
Method 1 – phase encoding ........................................................................32
2.2.2
Method 2 – canonical BOLD fitting (GLM) ...................................................34
2.2.3
Method 3 – deconvoluting BOLD-response functions .................................36
2.2.3.1
Block design vs. event related design......................................................36
2.2.3.2
Weighted Maximum Likelihood Estimate (WMLE) ...................................39
3 EXPERIMENT 1- Retinotopic mapping .......................................................... 41
3.1
Introduction ...................................................................................................41
3.1.1
Retinotopic mapping....................................................................................42
3.1.1.1
Eccentricity ..............................................................................................42
3.1.1.2
Polar angle...............................................................................................43
3.2
Methods .........................................................................................................43
3.2.1
FMRI measurements of retinotopy ..............................................................43
3.2.1.1
MR-protocols: ..........................................................................................44
3.2.2
MT+ localization ..........................................................................................45
3.2.2.1
FMRI measurements of MT+ ...................................................................46
3.2.3
Cortical flat maps.........................................................................................46
3.3
Results ...........................................................................................................47
3.3.1
Polar angle measurements..........................................................................47
3.3.2
Eccentricity measurements .........................................................................49
3.3.3
MT+ retinotopy and localizer measurements...............................................51
3.4
Discussion.....................................................................................................55
4 EXPERIMENT 2- Speedtuning ........................................................................ 58
4.1
Introduction ...................................................................................................58
4.1.1
Luminance - detection vs. discrimination.....................................................58
4.1.2
Isoluminance - detection vs. discrimination .................................................58
4.1.3
Two motion mechanisms.............................................................................59
4.2
Methods .........................................................................................................60
4.2.1
Subjects.......................................................................................................60
4.2.2
Visual display ..............................................................................................61
4.2.3
FMRI protocol..............................................................................................61
4.2.4
Stimulus & experimental design ..................................................................61
4.2.5
Analysis .......................................................................................................63
4.3
Results ...........................................................................................................63
4.3.1
Analysis of variance ....................................................................................65
4.3.1.1
Global BOLD-Analysis .............................................................................65
4.3.1.2
Analysis by velocity..................................................................................65
4.3.1.3
Analysis by brain area..............................................................................66
4.3.1.4
Velocity effects by cone-contrast direction...............................................67
4.3.2
Cluster analysis ...........................................................................................68
4.3.2.1
Stimuli for best separating hV4 and MT+.................................................69
4.4
Discussion.....................................................................................................70
4.4.1
Two motion pathways in human cortex .......................................................70
4.4.2
Color opponency in visual cortex.................................................................72
4.4.3
Colortuning ..................................................................................................73
4.4.4
Speedtuning ................................................................................................76
5 EXPERIMENT 3 - Colortuning......................................................................... 79
5.1
Introduction ...................................................................................................79
5.2
Methods .........................................................................................................82
5.2.1
Subjects.......................................................................................................82
5.2.2
Stimulus & experimental design ..................................................................82
5.2.3
MR-protocol.................................................................................................85
5.2.4
Fitting contrast response curve ...................................................................85
5.3
Results ...........................................................................................................86
5.3.1
Global effects of color, contrast and velocity ...............................................87
5.3.2
Interactions..................................................................................................87
5.3.3
Contrast response curves............................................................................88
5.3.3.1
Effects of stimulus contrast and velocity in V1 .........................................88
5.3.3.2
Effects of stimulus contrast and velocity in hV4 & V3A ............................89
5.3.3.3
Effects of stimulus contrast and velocity in MT+ ......................................92
5.3.4
Comparison at equal cone-contrast.............................................................93
5.3.4.1
Effects of stimulus velocity and color .......................................................94
5.3.5
Ellipse fits ....................................................................................................98
5.3.5.1
Effects of stimulus contrast, color and velocity ........................................98
5.4
Discussion...................................................................................................103
5.4.1
Neurophysiological studies and speed in V1 .............................................103
5.4.2
Neurophysiological studies and color in V1 ...............................................106
5.4.3
Neurophysiological studies and extra-striate cortex ..................................109
5.4.3.1
MT+ .......................................................................................................110
5.4.3.2
V3A & hV4 .............................................................................................111
5.4.4
Striate vs. extra-striate cortex....................................................................112
6 GENERAL DISCUSSION................................................................................ 114
7 APPENDIX ...................................................................................................... 118
7.1
Experiment 2 – Speedtuning......................................................................118
7.2
Experiment 3 - Colortuning ........................................................................120
7.2.1
ANOVA tables ...........................................................................................120
7.2.2
T-test tables...............................................................................................121
8 CITATIONS ..................................................................................................... 124
Abbreviations
Alpha
Flip angle
ANOVA
Analysis of variance
BD-path
Blob dominated path
BOLD
Blood oxygenation level dependent
Bk/Wh
Black/White
B/Y
Blue/Yellow
CO
Cytochromoxidase
CGC
Contrast gain control
EPI
Echo planar imaging
FLASH
Fast low angled shot sequence
FoV
Field of view
fMRI
Functional magnetic resonance imaging
FSPGR
Fast spoiled gradient echo sequence
FWHM
Full width half maximum
GLM
General linear model
HBM
Heterochromatic brightness matching
HFM
Heterochromatic flicker photometry
HRF
Hemodynamic response function
hV4
Human visual area 4
ID-path
Interblob dominated path
ISI
Inter stimulus interval
ITS
Inferior temporal sulcus
K-path
Koniocellular path
L-cones
Longwave sensitive retinal receptors
LGN
Laterale geniculate nucleus
M-cones
Middlewave sensitive retinal receptors
MD-path
Magnocellular dominated path
MEG
Magnetoencephalographic Imaging
M-path
Magnocellular path
MP-RAGE
Magnetization prepared rapid gradient echo sequence
MRI
Magnetic resonance imaging
MST
Medial superior temporal region (monkey)
MT
Medial temporal region (monkey)
MT+
Medial temporal complex (human)
PET
Positron emission tomography
P-path
Parvocellular path
rANOVA
Repeated measures analysis of variance
R/G
Red/Green
RF
Radio frequency pulse
RMS
Root mean square
ROI
Region of interest
S-cones
Shortwave sensitive retinal receptors
SET1
Spin echo T1-sequence
SPM
Statistical parametric mapping
TE
Time to echo
TFT
Thin-film transistor (display)
TI
Time for inversion
TR
Repetition time
V1
Primary visual cortex, striate cortex, Brodmans area 17
V2d/v
Visual area 2 dorsal/ventral
V3d/v
Visual area 3 dorsal/ventral
V3A
Visual area 3 accessory (a dorsal region)
V7
Visual area 7 (a dorsal region)
VO-1
Ventral occipital region 1
VO-2
Ventral occipital region 2
WMLE
Weighted maximum likelihood estimate
Summary
An organisms’ ability to perceive color and movement in its natural environment
greatly facilitates orientation and thus survival. Our understanding of cortical mechanisms
leading to color and motion perception is still at an early stage. It is known, that light sensitive
receptors in the eye give rise to several separate and spatially organized, cortical
representations of the visual scene in primate cortex. Extensive monkey research has shown
that these brain regions have different functional specializations. In humans, however, their
functional properties are still not well understood.
In three experiments multiple retinotopically mapped visual areas in humans were
investigated with functional magnetic resonance imaging (fMRI) regarding their characteristic
responses to stimuli varying in color, contrast and velocity. The measurements presented
quantify BOLD-responses to chromatic stimulation along thirteen directions in cone-contrastspace at four cone-contrast levels and at high and low stimulus velocity. They constitute the
most complete characterization of the chromatic response properties in human visual cortex to
date. On the basis of their response characteristics to these stimulus combinations,
retinotopically mapped visual areas will be associated to two psychophysically determined
motion perception mechanisms differing in chromatic sensitivity and temporal characteristics.
The lateral region MT+ is suggested to account for a fast motion mechanism, responding well
to all stimulus velocities and exhibiting some chromatic sensitivity, while dorsal area V3A
and ventral area hV4 are mediating the slow pathway by showing a similarly high sensitivity
to color and a preference for slowly moving stimuli. Velocity tuning in primary visual cortex
was found to be highly contrast dependent as speed preference in V1 reversed from low to
high speeds with increasing stimulus contrast. Evidence for cortical color-opponent
mechanisms in several regions is provided and stimulus compositions that allow the best
functional separation of dorsal and ventral visual areas are suggested.
In order to facilitate our understanding of cortical computation, this work provides
insight into how chromatic and motion signals are transformed as visual information is
transmitted from primary visual cortex to extra-striate regions. It is shown that color and
motion are processed over several levels of the cortical hierarchy while no stimulus property
is the exclusive domain of either processing stage.
GENERAL INTRODUCTION
A short history of color science
1 GENERAL INTRODUCTION
1.1 A short history of color science
`The rays to speak properly are not coloured´. With this statement Sir Isaac Newton
emphasizes color as a result of human perception and not a distinct property of the light
emitted by an object. He arrived at this conclusion after showing that white day light can be
decomposed into rays of different wavelengths, creating a spectrum, and that the color of an
object is determined by its surface properties and illumination conditions rather than a
physical property. His work Optiks (Newton, 1704) built the foundation for modern color
science.
The theory of trichromacy, first described by George Palmer in 1777 (Palmer, 1777),
rediscovered by Sir Thomas Young in 1802 (Young, 1802) and modified by Maxwell
(Maxwell, 1855) and Helmholtz (Helmholtz, 1867/1925) constitutes the first scientific theory
of color vision and perception. It is based upon color matching experiments carried out by
Maxwell. Maxwell's experiments demonstrated that most colors can be matched by
superimposing light from three sources with different colors, known as primaries. This
process is well known as additive color mixing. The theory simply states that there are three
types of color receptors, each sensitive in a different way to various wavelengths of light
(Figure 1.1). One receptor is sensitive primarily to blue light, another primarily to green, and a
third primarily to red. The appearance of a given mixture of light may then be described as a
point in a three-dimensional space. ´Brightness´ is defined as the sum of the responses of the
three channels, and ´hue´ and ´saturation´ are calculated in a ´color triangle´. Thus, the
fundamental dimensions of the Young-Helmholtz model of color perception are the responses
of three color-channels: Blue (B), Green (G) and Red (R). All other colors are described as
combinations of these three channels (B, G, R). The Young-Helmholtz formulation implies
that perceived color is a point property, which means color appearance is entirely determined
by the mixture of different wavelengths of light incident upon a single point on the retina.
Unfortunately, contrast phenomena show, that this is not the case. For example, the color
percept of a red region against a green background differs from the percept of the same red
region shown against a gray background. This demonstrates that the perceived color at a point
is affected by more than just the quality of the light at that particular point. A theory of color
appearance must thus take the spectral information of the spatial surround into account.
Also, the dependence of hue on brightness must be explained, namely that for instance bluegreens or blue-reds will shift towards blue as brightness increases. In the basic Young1
GENERAL INTRODUCTION
A short history of color science
Helmholtz color space, hue is the vector from the origin to the point in B, G, R space. If
brightness is changed, then the ratios between B, G, and R remain the same. Therefore, the
prediction is that the hue remains the same. However, as brightness increases, the perceived
hue tends towards either blue or yellow, known as the Bezold-Brücke-effect (Betzold 1873,
1874; Ejima, 1984). Taken together, these difficulties casted doubts upon the complete
fidelity of Young-Helmholtz color descriptions to human color perception because it is only
valid under certain conditions.
Motivated in part by the fact that there are no greenish-red or bluish-yellow colors as
well as the existence of color-after effects, Edwald Hering suggested a somewhat different
theory of color perception (Hering, 1878/1964). This theory postulates the existence of three
sets of paired variables, Yellow/Blue (Y-B), Red/Green (R-G), and Black/White (Wh-Bk).
These variables represent the responses of three "Opponent Processes". Thus, the sensation of
blue exists only at the expense of the sensation of yellow. Similarly, red and green are
opposing forces, as are white and black. The system is similar to the Young-Helmholtz model
in that color sensations are described in terms of a three-dimensional space. However, the
dimensions of this space are different. Hering's theory was rejected by many investigators
because he seemed to postulate the existence of chemical receptors which were affected in
opposite ways by different wavelengths of light (Kaufman, 1974). This mechanism was not
considered feasible, and there was no evidence for neural mechanisms, which could account
for the behavior of Hering's opponent-processes. However, Hering never claimed that the
opponent processes were at the receptor level, and later neurophysiological work showed that
such processes indeed exist (Hurvich & Jameson, 1957).
A unification of the two theories described above became known as the Dual Process
Theory. This theory was foreshadowed by von Kries (Kries von, 1905) and advanced by
Müller and Schrödinger in the early 1920s (Müller, 1924). A complete theoretical synthesis
and quantification of Hering's theory occurred not earlier than 1957 by Hurvich und Jameson
(Hurvich & Jameson, 1957)
Their "opponent-colors" theory is the basis for interpretation and stimulus generation
in the physiologically motivated color sciences (Hurvich & Jameson, 1955; Hurvich, 1956;
Jameson, 1956 ; Hurvich & Jameson, 1957). In order to account for above mentioned
surround effects, the opponent-colors model postulates an induction process whereby the
response at a given point is influenced by the responses at neighboring points (color-contrast).
If a stimulus produces a "red" response at some point, then it also produces a "green" response
in the surrounding points. Similarly, "blue" responses are accompanied by "yellow" responses
in the surrounding area. This mechanism would explain the observation that a red region
2
GENERAL INTRODUCTION
A short history of color science
against a green background appears "redder" than the same region surrounded by a white
background. This phenomenon is a direct analogy to the familiar edge effects in black and
white scenes or Mach-Bands. The point is that the appropriate choice of dimensions (R-G, YB, Wh-Bk) permits to explain both the chromatic and achromatic contrast effects by the same
computational principle, working independently in the three dimensions (Sloan & Bajcsy,
1975). Thus, it seems that the Hering theory of opponent processes, as extended and
quantified by Hurvich and Jameson, can account for the difficulties that the Young-Helmholtz
model faces.
The last major theory of color and color perception which will be reviewed is that of
Edwin Land and colleagues (Land & McCann, 1971 ). While experimenting with color
separation photographs (black and white pictures taken through red, green, and blue filters),
Land discovered that, under certain conditions, only two colors need to be mixed in order to
give the impression of a full color picture. Starting from this observation Land proceeded to
experiment with different combinations of filters and varying viewing conditions (Land,
1959). As a result of these experiments, Land has proposed the "Retinex" theory of color
perception (Land, 1977). The retinex theory proposes three separate "retinexes", each
sensitive to a different segment of the visual spectrum. Within each retinex there is a
representation of every object in the scene, rank-ordered according to its relative "lightness 1 "
within the retinex. The sensations of color are then described as combinations of these rank
orderings. According to Land, it is necessary, that the scene exhibits some randomness in
order for the effect to hold. When the scene is relatively uniform, the usual laws of color
mixture as described by Newton are observed. According to other researchers, the effects
Land demonstrates so vividly are nothing more than color contrast effects, combined with the
usual color mixture observations (Hurvich & Jameson, 1957). The Land theory is
unsatisfactory because of its explicit reliance upon completely global determination of the
color perception at each point. Another difficulty is that Land requires a judgment of
"lightness" independent of the illumination. With the right assumptions, this judgment can be
made (Horn, 1974). However, the necessary mechanisms are remarkably similar to the
induction process described by the opponent-colors model. However, the Land system does
provide a method for predicting the effect of two color mixtures, and has been used in the
design of a color display system (Panigrahi, 1973).
It is known to us today that both processing steps, photon count by means of
wavelength sensitive light receptors as well as opponent color computations, take place as
1
The quality or condition of being illuminated. The perceptual impression of luminance is called lightness It
refers to the brightness, or intensity value, of an object.
3
GENERAL INTRODUCTION
Neurophysiological basics
early as in retinal circuits. It is worth noting nonetheless, that above mentioned scientists came
to their conclusions without the physiological knowledge available today, which highlights
their extraordinary feat.
The following section reviews contemporary neurophysiological knowledge about the
visual system. An introductory overview on the anatomy and function of the visual system is
given and subsequent sections will describe the consecutive stages of visual information
processing in more detail with emphasis on chromatic signal transduction. The chapter closes
with the specification of experimental objectives studied in this thesis.
1.2 Neurophysiological basics
1.2.1 Anatomy and function of the visual system
The sheer number and size of the cortical areas that participate in visual information
processing indicates the importance of primate visual perception. Approximately 60% of the
entire neocortex take part in perception, interpretation and reaction to visual information.
More than 30 distinct and highly interconnected visual areas could be described in the primate
cerebral cortex (Felleman & Van Essen, 1991; Van Essen et al., 1992; Merigan, 1993; Van
Essen & Gallant, 1994; Salin & Bullier, 1995).
The path of the visual information processing begins with light absorption in the
photoreceptors of the retina. Within the retina a large part of visual processing is conducted by
means of these light receptors, bipolar, amacrin, horizontal and ganglion cells. The visual
information is transmitted via the optic nerve through the optic chiasm and, by means of the
optic tract, transferred to the lateral geniculate nucleus (LGN) in the thalamus. The LGN is
considered the main relay station for visual information from retina to cortex. Approximately
90% of the retinal ganglion cells project to the LGN, while the remainders connect to the
superior colliculi and the pulvinar. Each thalamic hemisphere contains one LGN, which is
distinguishable in 6 layers. The two inferior layers consist of relatively large magnocellular
neurons, while the four superior layers contain relatively small parvocellular neurons. In
between the parvo- and magno-cellular layers reside the koniocellular layers. Because retinal
fibers of the nasal retina cross over in the optic chiasm, each LGN receives visual information
stemming from the contra-lateral visual field. For instance, the left LGN receives information
from the left halves of each retina, which sample the right visual field. Each individual cell
layer in the LGN receives signals exclusively from one eye. The fibers of the nasal
contralateral retina project to the layers 1, 4 and 6, while the temporal ipsilateral fibers project
to layers 2, 3 and 5 of the LGN. The optical radiation forwards the signals from LGN towards
4
GENERAL INTRODUCTION
Neurophysiological basics
the primary visual cortex (V1), situated in the calcarine sulcus of the occipital lobe of
primates. Primary visual cortex is organized in layers and columns of distinct cell types, while
specific color-responsive cells were found in blob-structures known as cytochromoxidase rich
blobs (CO-blobs). The energy consumption of these cells could be functionally marked using
a technique called cytochromoxidase-staining. After specific stimulus classes have been
presented, for instance color, this method generates stained neuronal regions, which have been
active during the stimulus presentation. Inactive regions will not be stained with
cytochromoxidase leading to the interpretation that active regions functionally contribute to
the processing of the presented stimulus because their energy consumption is elevated during
the respective experimental stimulation.
Commencing in V1, the cortical processing of visual information is believed to be
separated into two main paths (Ungerleider & Mishkin, 1982; Goodale & Milner, 1992).
Firstly, the parietal path, projecting into parietal regions of cortex and secondly the temporal
path projecting towards inferior temporal cortex. The parietal stream is assumed to be mostly
associated with spatio-temporal information processing associated with an ‘action’-system
(the WHERE-path), while the temporal stream is believed to be mainly associated with
recognition of objects, a perceptual system, and thus with color-, pattern- and form-perception
(the WHAT-path) (Ungerleider & Mishkin, 1982; Goodale & Milner, 1992).
While the current knowledge of early visual processing, starting from the retina to
primary visual cortex, can be regarded as quite substantial, knowledge of higher cortical
processing may be best described as insufficient. If and to what sort of stimuli the higher
cortical areas are specialized as well as response properties to ‘simple’ stimulus attributes
such as color are still a matter of debate. This thesis attempts to characterize the response
properties of several visual areas located in human visual cortex in the color, motion and
contrast domain. The next sections describe the path of visual information as it is transmitted
in consecutive stages in retino-geniculo-cortical pathways in more detail.
1.2.1.1 Three receptor classes
Light-sensitive photoreceptors in the retina constitute the first processing stage of the
visual information. Under photopic viewing conditions (at high light intensity levels)
receptors, called the cones, are receptive, while scotopic vision (at low light intensity levels),
is maintained by the rods. In a normal human retina three classes of cones are found. Each of
which contains a slightly different light absorbing pigment. They can be considered the
physiological implementation of Helmholtz’ theory of trichromacy. The absorption spectra of
5
GENERAL INTRODUCTION
Neurophysiological basics
each class of cones are known (Brown & Wald, 1964; Marks et al., 1964) and can be
described by three overlapping functions (Figure 1.1).
S
M
L
Figure 1.1: Normalized cone spectral sensitivities
Lines: blue (short-wavelength sensitive receptor; S-cones), green (middle-wavelength sensitive
receptor; M-cones), red (long-wavelength sensitive receptor; L-cones).
The short-wave- or S-cones maximally absorb light at ~440 nm, the middle-wave- or
M-cones maximally absorb light at ~530 nm and the long-wave- or L-cones maximally absorb
light at ~560nm (Schnapf et al., 1987). The number and density of each cone class varies
across the retinal receptor mosaic (Roorda, 1999). The S-cones amount to 5-10% in the
human retina. The ratio of L- and M-cones is highly variable across observers and can range
from 0.82 to 9.71 (Brainard et al., 2000). The region of highest acuity in the human retina is
called the fovea, in which receptor density is highest. The central 0.1° of the fovea is made up
entirely of L- and M-cones. The amount of S-cones at 1.0° increases to 6%.
1.2.1.2 Neurophysiological implementation of color-opponency
The following section uses terms such as red, green, blue and yellow light. The
perceptual impressions of red, green and blue can however not be regarded equal to excitation
of the L-, M- or S-cones (see Figure 5.2). While the excitation of the L-cones yields a reddish
percept, excitation of the S-cones results in a dark violet. The notation to describe properties
6
GENERAL INTRODUCTION
Neurophysiological basics
of receptive fields in the retinal ganglion cells and LGN is conventional but can be
misleading.
Russel De Valois and colleagues, pioneered neurophysiological studies of coloropponency by showing that response patterns from color sensitive neurons in the LGN of the
macaque are not in concordance with Helmholtz’ theory of trichromacy (Helmholtz,
1867/1925), but are in line with the opponent process theory postulated by Hering (Hering,
1964; De Valois et al., 1966). Some cells were found to be excited by red light and inhibited
by green light (R+G-cells). Other cells were excited by green light and inhibited by red light
(G+R-cells). In addition, cells were found that could be excited by yellow light and inhibited
by blue light (Y+B- cells) as well as cells that were excited by blue light and inhibited by
yellow light (B+Y-cells). Another population of cells did not show spectral opponent behavior
whatsoever. These cells were excited by light containing a mixture of all wavelengths while
the absence of light inhibited them (Wh+Bk-cells). The opposite was also found, namely cells,
which show inhibition by stimulation with all wavelength light and excitation through absence
of all wavelength light (Bk+Wh-cells). Together these cells can be interpreted as neuronal
implementation of Herings opponent process theory. However, this kind of opponent
processes can already be found in the retinal bipolar- and ganglion cells (de Monasterio, 1982;
Lee et al., 1988; Lee, 1990).
The specific response pattern of R+G-cells for instance arises through combination of
excitatory input from L-cones with inhibitory input from M-cones. Essentially, output from
these cells is governed by the difference between L- and M-cones (L-M). The output from
G+R-cells is dominated by excitatory input from M-cones and inhibitory input from L-cones,
yielding a similar difference computation (M-L). The output of B+Y-cells is generated through
excitatory input from S-cones and the inhibitory sum of L- and M-cones (S-(L+M) = S-L-M),
while the output from Y+B-cells is produced by the excitatory sum of L- and M-cones opposed
by an inhibitory signal from the S-cones ((M+L)-S = M+L-S). Output from Wh+Bk-cells
originates in the excitatory sum of L-, M- and S-cones (L+M+S), while the output from
Bk+Wh-cells is created through inhibitory signals from all cone classes (-L-M-S).
Knowledge of these fundamental neural computations will govern the stimulus
generation for all experiments in this thesis. Controlled excitation of retinal photoreceptors
and downstream neural pathways, in multiple directions of color-space is essential to achieve
the major goal of this work, which is the characterization of chromatic responses in multiple
cortical visual areas.
7
GENERAL INTRODUCTION
Neurophysiological basics
1.2.2 Spatial response properties of visual neurons
1.2.2.1 Receptive fields of retinal ganglion cells
Each retinal ganglion cell receives visual information from upstream-connected
photoreceptors, which in turn respond to stimuli in designated positions in the visual field.
The location in the visual field in which stimuli affect neuronal activity levels is called the
receptive field of that neuron. Retinal ganglion cells have small concentric receptive fields
with an antagonistic center-surround organization (Lee, 1996). An ON-center ganglion cell
starts firing, when a light source stimulates exactly the center of its receptive field. The firing
of the ganglion cell is inhibited, when the antagonistic surround is stimulated exclusively.
Similarly, OFF-center ganglion cells exist, for which stimulation of the center causes the
highest inhibition, while stimulation of the surround yields an excitation. If, however the
entire receptive field is illuminated equally, a relatively weak response can be expected. The
antagonistic center-surround organization of retinal ganglion cells causes a high sensitivity for
local contrast changes, while equal illumination changes across the entire visual field have
little effect on these cells firing rate.
1.2.2.2 Receptive fields of primary visual cortex (V1)
The response properties of neurons in primary visual cortex are more complex than in
the retina or the LGN. V1 neurons do not respond very well to point-like stimulation, while
responding vigorously to stripes of light. Ground-breaking work on this matter has been
presented by David Hubel and Torsten Wiesel using cats and monkeys (Hubel, 1959). They
distinguished three cortical types of neurons, depending on their preferred type of stimulation.
Simple cells respond to bars of light with a particular orientation. The receptive fields
of simple cells are elongated, oriented in a certain direction in space and divided into
excitatory and inhibitory zones. A strong response from such a neuron can be expected when
a stripe-like stimulus or contrast edge is presented to its receptive field, which is equal in
orientation and size to its excitatory zone.
Complex cells exhibit no clear distinction between an excitatory and inhibitory zone.
They also selectively respond to oriented stripe-like or contrast edge stimuli, but the exact
position within the receptive field is irrelevant. As opposed to simple cells, complex cells
prefer edges independent of contrast direction. Additionally they respond selectively to
stimuli moving over their receptive field in specific directions.
Hypercomplex cells respond to stripes and angles of a certain length, which move in
the preferred direction across its receptive field. The elongated excitatory and inhibitory zones
8
GENERAL INTRODUCTION
Neurophysiological basics
of simple and complex cell receptive fields are thought to emerge through combination of
several concentrically organized cells (Hubel, 1962).
Neurons with distinct receptive field attributes are organized in columns. Their
properties can be differentiated by position of the receptive field (which is fairly equal within
one column), ocular-dominance and orientation of the receptive fields. A column representing
a certain position in the visual field is called a hyper column and contains two ocular
dominance columns and numerous orientation columns (0°-360°).
1.2.2.3 Retinotopic organization
Adjacent retinal ganglion cells have adjacent and overlapping receptive fields and
project to neighboring neurons of the next stage of visual information processing.
Consequently, neighboring locations in the visual field are represented in neighboring neurons
in cortex. This organizational principle, by which the spatial arrangement of retinal ganglion
cells is preserved, is called retinotopy. This spatial organization is maintained in the LGN and
cortex until the receptive fields of higher tier areas become progressively larger and spatial
information is integrated to abstract categories such as for instance faces and cars (Press et al.,
2001; Smith et al., 2001; Huk et al., 2002; Wade et al., 2002; Wandell et al., 2005; GrillSpector et al., 2006). The retinotopic organization of many visual areas permits measurements
of the cortical representation of the visual field efficiently and with high spatial resolution
(DeYoe et al., 1994; Engel et al., 1994; Sereno et al., 1995; Engel et al., 1997b). By means of
retinotopic mapping, several visual areas can be distinguished solely on their spatial
representation of the visual field and independent of any other functional criterion.
However some dispute regarding topographic organization of some higher visual areas
pertains in the field (Hadjikhani et al., 1998; Tootell, 2001; Wade et al., 2002; Brewer et al.,
2005). Different methods have yielded slightly different results and interpretations. The first
series of experiments in this thesis will present several retinotopic measurements, the results
of which will serve as a precise cortical location to generate regions of interest (ROIs)
representing distinct visual field maps. Subsequent experiments will use these ROIs to
investigate the functional properties pertaining to color, motion and contrast dependences
within separate visual areas.
9
GENERAL INTRODUCTION
Neurophysiological basics
1.2.3 Anatomy and function of geniculo-cortical pathways
1.2.3.1 Anatomy
Several distinct paths project from retina to cortex. The paths can be distinguished on
the basis of multiple anatomical criteria. They differ in morphology, density of corresponding
cells, connections within the retina and presumably independent projections. To date three
major pathways are discriminated and have been extensively investigated in the monkey
brain.
The parvocellular path (P-path) originates in the midget ganglion cells in the retina
(Dacey, 2000). It was named after the parvocellular neurons in LGN layers. These neurons
have ON- and OFF-center receptive fields (Dacey, 2003). In the inner fovea these centers
consist of only one L- or M-cone. Anatomical studies have shown that S-cones contribute
only to OFF-center ganglion cells (Klug, 1993). The midget ganglion cells project into the
four parvocellular layers of the LGN and from there mainly to layer 4Cβ (Chatterjee &
Callaway, 2003) but also to layers 4A and 6 of primary visual cortex (Lennie, 1988;
Fitzpatrick, 1996; Solomon & Lennie, 2005; Conway & Livingstone, 2006).
The magnocellular path (M-path) originates in the parasol cells (M-cells) which in
turn receive their input from diffuse bipolar cells (Dacey, 2000, 2003). The diffuse bipolar
cells obtain additive input from L- and M-cones (Calkins, 1996; Lee, 1996a; Dacey, 2000;
Jacoby, 2000; Dacey, 2003; Conway & Livingstone, 2006). The receptive fields of the Mcells have an antagonistic center-surround organization. These neurons project into the two
magnocellular layers of the LGN. They exhibit a high sensitivity for luminance contrast
(L+M). The magnocellular layers of LGN project mainly to layer 4Cα (Chatterjee 2003) but
also to layers 4B and 6 in primary visual cortex (Lennie, 1988; Fitzpatrick, 1996; Solomon &
Lennie, 2005).
The koniocellular path (K-path) originates in the small bistratified retinal ganglion
cells (Dacey, 1993; Hendry & Yoshioka, 1994). The receptive fields of these neurons consist
of an excitatory S-cone center and an inhibitory surround formed by additive L- and M-cone
signals (S-(L+M)). They project into the koniocellular layers of the LGN (Casagrande, 1994;
Hendry & Yoshioka, 1994) after which this path was named. These layers reside between the
magno- and parvocellular layers in the LGN. In primary visual cortex these cells are known as
Blue-ON-cells and can be found in layer 4A (Chatterjee & Callaway, 2003) but also in the
CO-rich blobs of the superficial layers (1,2,3) (Chatterjee, 2002; Chatterjee & Callaway,
2003; Sincich & Horton, 2005; Conway & Livingstone, 2006). In layer 4A, Chatterjee (2003)
also characterized Blue-OFF-cells which respond most vigorously to pure S-cone stimuli with
10
GENERAL INTRODUCTION
Neurophysiological basics
negative sign and exhibit weak responses to the sum of L- and M-cones (Chatterjee &
Callaway, 2003).
As early as in the primary visual cortex (V1) there are numerous anatomical
connections between the magno- and parvocellular streams (Merigan & Maunsell, 1993;
Yoshioka et al., 1993; Lund et al., 1995), originating in the LGN, while the interaction with
the koniocellular path are currently highly investigated (Sincich & Horton, 2005). Beyond V1,
in visual areas V2, V3 and V4, the connectedness between different, anatomically defined
compartments, increases substantially (Yoshioka et al., 1992; Levitt et al., 1994). Area V4,
for instance, was estimated to receive about an equal proportion of inputs originating from
magno- and parvocellular LGN-layers (Ferrera et al., 1994). Even motion sensitive region V5
(MT), which was long believed to solely receive magnocellular input (Maunsell et al., 1990)
could be shown, by means of neuroimaging, to receive contributions from other paths as well
(Wandell et al., 1999).
1.2.3.2 Functional specialization in the LGN
The anatomically distinct paths are believed to exhibit differential spatial and temporal
properties. P-cells for instance are most sensitive at low temporal flicker rates, while M-cells
are more sensitive to high flicker frequencies (Marrocco et al., 1982; Hicks et al., 1983; Van
Essen, 1990). In humans, the receptive field size of M-cells is up to ten times larger than the
receptive fields of P-cells (Van Essen, 1990) and in addition have a much lower retinal
distribution density as compared to P-cells (Perry, 1984; Grunert, 1993). K-path-neurons
show a low retinal distribution density as well (Dacey & Lee, 1994).
High spatial frequencies seem to be selectively preferred by the P-path. Highest spatial
resolution is however known to be inherent to achromatic viewing, associated with the Mpath, posing a profound inconsistency with respect to the specific task segregation between
M- and P-path. Different receptive field properties as well as the aforementioned anatomical
connections within each path, indicate differential specialization regarding transmission of
color information. The P-path seems to code color whereas the M-path codes luminance
(Spillmann & Werner, 1990). The spatial density and quantity of neurons in the P-path is
significantly higher than in any other retinal path. In the fovea for instance, each cone sends
signals to two neurons in the P-path (one ON- and one OFF-signal). In fact, the ganglion cell
mosaic of the P-path consists of 70% of the 1.25 million retinal ganglion cells (Rodieck,
1998). This suggests the P-path to be responsible for the high spatial achromatic viewing,
even though it is thought to mainly transmit color information.
11
GENERAL INTRODUCTION
Neurophysiological basics
In summary, it can be stated that at least three anatomically distinguishable channels
exist (P-, M-, K-path), that carry visual information from retina to cortex. Several lines of
evidence suggest that the anatomical segregation as present in the LGN may not be
maintained throughout cortex, as visual information from several paths may be merged in
order to achieve maximal integration (Nealey et al., 1991; Merigan & Maunsell, 1993; Ferrera
et al., 1994; Nassi & Callaway, 2006). The next section reviews relevant findings mainly in
the macaque brain.
1.2.3.3 Functionally specialized pathways in the cortex
Comparison of chromatic response patterns in human striate cortex to response
profiles in higher tier areas is a major goal of this thesis. The visual information is believed to
have reached the extra-striate regions by projections described in the next section.
In cortex, a clear segregation of the three subcortical pathways as found in the LGN
cannot be reported, because a blending of the paths is progressively evident as a multitude of
interconnections in V1 and higher areas occur (Sincich & Horton, 2005). However, several
groups postulate, that the three paths remain highly segregated as they advance into the extrastriate cortical hierarchy (DeYoe & Van Essen, 1985; Livingstone & Hubel, 1987; Roe &
Ts'o, 1999). Cytochromeoxidase-staining in the macaque has provided functional landmarks
in cortical regions such as CO-rich blobs in V1 and thin, thick and pale stripes in V2,
responsivity of which may be investigated regarding the contributions of the known
projections from the LGN.
The retino-geniculate M-path is continued as the magnocellularly dominated (MD)
path in cortex. This path has been identified as having the least convergence with the other
two pathways (DeYoe & Van Essen, 1985) and is mainly, but not exclusively, fed by
magnocellular projections. In V2, this path projects into the CO-thick stripes. Livingstone and
Hubel report an anatomical and physiological model, which proposes this path to be color
blind, highly contrast sensitive and having poor spatial resolution (Livingstone & Hubel,
1988).
The cortical pathway exhibiting highest convergence of all described projection routes
is the blob dominated (BD) path (Fitzpatrick et al., 1985). Part of this path are the CO-richblobs in V1 which appear to receive input from all three geniculo-striate paths (M-, P-, Kpath) (Fitzpatrick et al., 1985; Lachica et al., 1992; Yoshioka et al., 1993; Sincich & Horton,
2005). Experiments in which P-layers of the LGN were selectively deactivated, show
nonetheless, that many cells in the blobs could still be activated, presumably by cells that
combine M- and P-path information, showing residual magno-cellular responsivity (Nealey et
12
GENERAL INTRODUCTION
Neurophysiological basics
al., 1991). Livingstone and Hubels model assigns this path high color selectivity, low contrast
sensitivity and high spatial resolution (Livingstone & Hubel, 1987).
The third cortical path pertains to the CO-interblob regions in V1 and is called ID-path
(interblob dominated). The interblob regions receive substantial input from parvocellular
LGN-layers (Lachica et al., 1992; Yoshioka et al., 1993) but also from magnocellular layers
(Yoshioka et al., 1993). Selective deactivation of LGN-layers could also show that the
interblob neurons in V1 can be excited by parvo- and magnocellular contributions (Nealey et
al., 1991). The interblob outputs project to the CO-pale stripes in V2. Beyond V2, this path is
believed to supply the ID-path subregions of V4 and the posterior inferotemporal cortex
(Felleman et al., 1997a).
The selectivity of neurons comprising each path is characterized by response profiles
that seem sensitive to a variety of visual information (DeYoe & Van Essen, 1988; Peterhans
& von der Heydt, 1993). The MD-path is highly selective for binocular disparity, motion
direction, velocity, orientation and low spatial frequency. The BD-path is highly selective for
wavelength information and high spatial frequencies, while the ID-path selectively prefers
orientation, disparity, low spatial frequencies and wavelength information (Van Essen &
DeYoe, 1995).
1.2.3.4 Cortical color mechanisms
Color processing and characterization of distinct visual areas with respect to their
color sensitivity are actively debated subjects (Gegenfurtner, 1999).
As indicated above, early studies described the existence of color selective cells of the
macaque monkey (Hubel & Wiesel, 1968; Dow & Gouras, 1973; Gouras, 1974; Poggio et al.,
1975; Thorell et al., 1984). A turning point regarding our understanding of cortical color
mechanisms commenced by the work of Livingstone and Hubel in 1984 (Livingstone &
Hubel, 1984). They report conglomerates of non-oriented, monocular, color-selective cells in
CO-blobs of macaque V1. Within the CO-blobs however, an additional color cell type was
identified. Livingstone and Hubel had found the double-opponent cells in V1, which were
first found in the goldfish retina (Daw, 1968). These cells respond strongest when their
receptive field centers are stimulated with one color and their surround with the opponent
color (Daw, 1968; Hubel & Wiesel, 1968; Conway, 2001). Opponent mechanisms, red-green
for instance, are hence prevalent in the receptive fields’ center as well as in the surround of
these cells. Red-green double-opponent cells are thus able to compare the relative amounts of
red-green in one part of a scene with the amount of red-green in an adjacent part of the scene.
Hence they respond best to local color contrast.
13
GENERAL INTRODUCTION
Neurophysiological basics
Ts’o and Gilbert observed B/Y or R/G opponent-color mechanisms in the blobs of V1,
and specified additional cells with receptive fields having either a B/Y or R/G opponent color
mechanism solely in the center, while a broadband inhibitory surround was found (Ts'o &
Gilbert, 1988). They called this additional cell class modified Type II-cells (Ts'o & Gilbert,
1988). In addition, they discovered cells that were color selective and orientation sensitive.
These cells were most frequently found in between borders of blobs and interblobs. In the
same study they could show, that CO-blob color cells are organized in clusters with similar
sensitivity to opponent colors. Some CO-blobs are designated to respond to either R/G or B/Y
opponent colors.
However, as opposed to LGN neurons, V1 contains cells that are not exclusively
sensitive along the two cardinal color mechanisms (S-(L+M), L-M), but also respond
vigorously to mixed colors (Lennie et al., 1990; Solomon & Lennie, 2005; Conway &
Livingstone, 2006). This mixed color selectivity is also prevalent in macaque V2
(Gegenfurtner et al., 1996) and V3 (Gegenfurtner et al., 1997).
The functional organization and architecture in V2 is different as compared to V1
(Kiper et al., 1997). CO-staining in V2 revealed parallel thin, thick and pale stripes (DeYoe &
Van Essen, 1985). It was assumed that color information is primarily processed in the thin
stripes, motion information in the thick stripes and form information mostly in the pale
stripes. Gegenfurtner and colleagues could show, however, that neurons in macaque V2 are
not restricted to one stimulus attribute such as color, form or motion (Gegenfurtner et al.,
1996). In addition they report that color sensitive neurons are also not restricted to the thin
stripes but occur in other stripe classes which are known for different specializations as well
(Gegenfurtner & Hawken, 1996; Gegenfurtner et al., 1996). It became thus convincing that a
strict segregation regarding stimulus attributes in parallel processing streams is not prevalent
in V2. Combined processing of stimulus attributes (multiplexing) in V2 is more likely.
Similar results could be obtained for area V3 in the macaque (Gegenfurtner et al.,
1997). It was found that V3 receives strong inputs, originating from both, the magno- and the
parvo-path, and has prominent projections to areas MT and V4, which are believed to be the
cortical motion or color center respectively. Area V3 may thus represent an important cortical
site for integrating visual signals, originating in anatomically different pathways. Orientation
selectivity in V3 was prevalent in 80% of the cells, similar to V2 (Gegenfurtner et al., 1996;
Gegenfurtner et al., 1997). However, neurons in V3 seem to prefer lower spatial and higher
temporal frequencies than V2 neurons (Gegenfurtner et al., 1997). Contrast thresholds
obtained from V3 neurons were extremely low. Approximately 40% of all neurons showed
strong selectivity for the direction of a moving stimulus. However, no direction selective cells
14
GENERAL INTRODUCTION
Neurophysiological basics
were found in layer 4 of V3, which constitutes the input termination site for V1 and V2
projections. This finding indicates, that selectivity for the direction of motion may develop
within V3 (Gegenfurtner et al., 1997).
In the macaque monkey the proportion of color selective cells was found to be
approximately equal in V2 (50%) and V3 (54%) (Gegenfurtner et al., 1996; Gegenfurtner et
al., 1997). In addition it was reported that V3 subpopulations, sensitive to color or motion, are
largely overlapping. Cells, responding reliably to directional signals elicited by drifting
isoluminant chromatic gratings were also found (Gegenfurtner et al., 1997). In summary,
these neurophysiological results indicate, that there is a significant interaction between color
and motion processing in macaque area V3.
The motion sensitive area MT, situated at the lateral medio-temporal regions in the
macaque is classically believed to receive predominately magnocellular input (Maunsell et al.,
1990) and was suggested to completely lack color selectivity (Zeki, 1983c). Using a
combination of physiological and psychophysical techniques, subsequent studies were able to
show that color information does contribute to the response of MT neurons in monkeys
(Dobkins & Albright, 1994; Croner & Albright, 1999; Seidemann et al., 1999; Thiele et al.,
1999). Moreover, Gegenfurtner et al. (1994) showed that most MT neurons do respond to
chromatic modulation, albeit much less vigorously than to luminance variations (Gegenfurtner
et al., 1994). Recent imaging reports on human subjects indicate that MT+ as a whole gives
responses to chromatically defined stimuli, even though it may contain few, if any, coloropponent cells (Tootell et al., 1995b; Wandell et al., 1999; Liu & Wandell, 2005). The ventral
occipital region hV4 is described in detail in the next section.
1.2.3.5 Cortical color center ?
The delineation of a cortical center, specialized for a perceptual property such as color,
implies that the perception of this attribute is represented through activity of neurons within
this cortical region (Zeki, 1990).
Early indications for the existence of an unequivocal color center originated from
single cell investigation of the macaque area V4. It has been argued that V4 is a specialized
color center because it contained solely color selective cells (Zeki, 1983b, 1983a, 1983c,
1983d). First estimates proposed 100% color sensitive cells in V4 (Zeki, 1973). Subsequent
studies found substantially less color selective cells. Schein et al. did not find more than 20%
(Schein et al., 1982). Even though the multitude of studies show little congruence, it is highly
likely, that V4 contains the same amount of color selective cells as V1, V2 or V3. Schein and
Desimone showed that most V4 neurons do not show higher sensitivity with regard to
15
GENERAL INTRODUCTION
Neurophysiological basics
wavelength-selectivity compared to retinal color-opponent ganglion cells or parvocellular
LGN-cells (Schein & Desimone, 1990).
A main argument against a clear-cut color area V4, provide lesion studies. Lesions in
V4 of the macaque do not exclusively cause deficits in color perception (Heywood & Cowey,
1987; Schiller, 1993) but include also impairments of other stimulus attributes such as form
discrimination (Walsh et al., 1992; Merigan, 1996), object recognition (Schiller, 1995;
Merigan & Pham, 1998; Walsh et al., 2000), texture discrimination (Merigan, 2000) and
focusing of attention (De Weerd et al., 1999).
In humans, it is long known that bilateral lesions in parts of the fusiform and lingual
gyrus on the ventro-medial side of the occipital-temporal cortex (see Figure 3.1.) can cause
severe deficits in color perception (Meadows, 1974; Zeki, 1990). For that reason this region
on the ventral cortical surface has been assigned the human homologue to macaque V4. Using
modern imaging techniques it could be shown that the ventral cortical surface is highly active
during color perception (Lueck et al., 1989; Corbetta et al., 1991; Gulyas et al., 1994;
Kleinschmidt et al., 1996; McKeefry & Zeki, 1997; Wade et al., 2002; Brewer et al., 2005;
Liu & Wandell, 2005). Hadjikhani (1998), using functional magnetic resonance imaging, also
found a region in ventral occipital cortex, which was highly activated by colored stimuli
(Hadjikhani et al., 1998). They argued that this region is not the homologue to macaque V4
and named it V8 (Hadjikhani et al., 1998). It is however very likely that V8 is not a new
separate color area, solely responsible for color perception, as could be shown in a study,
investigating the topographic representation of the visual field along the ventral surface
(Wade et al., 2002). Wade (2002) found activity elicited through colored stimuli extending
from V1 along the ventral surface, showing that V8, if it exists, is not the only region
responding strongly to color.
Additional primate studies reported that V4 neurons, similar to neurons in V2, can be
selectively excited by certain lengths, widths, orientations and forms of the stimuli used
(Desimone & Schein, 1987). V4 seems to feature a multitude of processes besides coding
color. An additional indicator against V4 as a clear-cut color center are studies which showed
that V4 neurons can selectively respond to direction of motion and can also be modulated by
attention processes (Moran & Desimone, 1985; Tolias et al., 2005a).
In summary, it is highly unlikely that a single cortical region exists which responds
exclusively to colored stimuli. It is thus more probable, that color perception is the results of
simultaneous neural activity in several cortical sites and that color-coding takes place in
several visual areas simultaneously.
16
GENERAL INTRODUCTION
Experimental objectives
1.2.3.6 Chromatic responses in human cortex
As indicated above, most of the studies that described the existence of color selective
cells have been conducted using single cell recordings in the macaque monkey (Hubel &
Wiesel, 1968; Dow & Gouras, 1973; Gouras, 1974; Poggio et al., 1975; Thorell et al., 1984).
Compared to the amount of single cell monkey-studies, only few noninvasive human fMRIstudies regarding this field exist (Kleinschmidt et al., 1996; Engel et al., 1997b; McKeefry &
Zeki, 1997; Hadjikhani et al., 1998; Zeki & Marini, 1998; Beauchamp et al., 1999; Bartels &
Zeki, 2000; Engel & Furmanski, 2001; Wade et al., 2002; Liu & Wandell, 2005). Some PETstudies were also conducted (Lueck et al., 1989; Zeki, 1991). The goal in several of these
studies was to measure activity, elicited by colored stimulation, aiming for the cortical color
center (Lueck et al., 1989; Zeki et al., 1991; Beauchamp et al., 1999; Bartels & Zeki, 2000).
The subtraction methodology was used, which implies the acquisition of functional brain
activity, while the subjects view either colored or black/white stimuli in order to subtract the
measured achromatic- from the chromatic activity, to be left with cortical regions responding
stronger to color than to achromatic stimulation. While this procedure did in fact reveal the
ventral cortical surface to show higher activity to chromatic than to achromatic stimuli,
conclusions regarding opponent-color mechanisms as described above (see section 1.2.1.2.)
cannot be made. If the stimuli are not equated in a way, which takes cone-computations,
known to be inherent to retino-geniculo-cortical pathways into account, conclusions beyond a
comparison of chromatic versus achromatic responses are not feasible. In addition, how and to
what extend distinct visual areas respond to chromatic or achromatic contrast modulation
could not be derived because retinotopic mapping was not performed in most of the early
studies (Kleinschmidt et al., 1996; McKeefry & Zeki, 1997; Zeki & Marini, 1998; Bartels &
Zeki, 2000). A detailed description of chromatic stimulus control is given in section 2.1.2.
Studies investigating chromatic responses in human visual cortex using stimuli controlled for
cone excitation will motivate the experimental objectives of this thesis and will be introduced
in the next section.
1.3 Experimental objectives
The main goal of this thesis is to characterize retinotopically defined human visual areas
regarding their differential response patterns to chromatically defined stimuli. A second
objective is to assess the dependence of the chromatically induced response patterns to
additional stimulus attributes, such as stimulus velocity and contrast. This is of interest
because these measurements would allow the assessment as to how the visual information is
17
GENERAL INTRODUCTION
Experimental objectives
transformed on its way from primary visual cortex to extra-striate regions and what particular
stimulus attribute combinations may differentially be preferred in distinct visual areas.
In order to study in what way chromatic information, originating in distinct retinogeniculo pathways, is differentially processed in cortex, it is crucial to apply physiological
stimulus criteria rather than perceptual (see section 1.2.3.6). It is possible to construct
chromatic stimulation that emphasizes activity in neurons of either of these pathways, while
modulations of other pathways can be kept constant. To achieve this, one must know the
chromatic properties of the first input stage to these pathways, which constitute the spectral
sensitivities of three light receptor classes in the retina (see section 1.2.1.1). The experimental
color-conditions can thus be controlled in terms of cone excitation. How these cone-signals
are processed in post-receptoral mechanisms (see section 1.2.1.2 and 2.1.2.1) should be taken
into account, to obtain comparable and conclusive results pertaining to these mechanisms. In
order to achieve equitable comparison between the response properties in several visual areas,
originating in separate pathways, the presented spectral stimulus energy should be computed
equal with regard to the initial encoding stage. Comparison of response patterns generated at
equal-cone-contrast allows for conclusions regarding selectivity and chromatic preference
across visual areas.
It seems that from previous fMRI studies investigating chromatic responses in human
cortex, only a minority used cone-contrast equated stimuli (Kleinschmidt et al., 1996; Engel et
al., 1997b; Engel & Furmanski, 2001; Wade et al., 2002; Brewer et al., 2005; Liu & Wandell,
2005). Nonetheless, as this group of studies showed, it is possible to vary the stimulus conecontrast in stages (parametric design) to asses, in what way retinotopically defined visual
areas differ in their response characteristics depending on cone-contrast sensitivity.
An influential fMRI study showed, that if cone-contrast stimulation is applied, it is
possible to yield meaningful results regarding chromatic sensitivity in V1 and V2 (Engel et
al., 1997b). Engel (1997) presented stimuli at several parameterized cone-contrasts levels and
several directions in cone-contrast space. The cone-contrast space is a three-dimensional
physiologically based color space, in which the three retinal receptor-excitations constitute the
axes. For a detailed description of this and other color spaces see section 2.1.1.1. Engel (1997)
could show that V1 and V2 respond vigorously to stimuli along the L-M and S-(L+M) axis in
cone-contrast space and that temporal flicker frequency may have a differential effect on
responses elicited by different combinations of cone excitations.
The chromatic stimuli presented by Engel and colleagues were constructed to sample
responses to cone-contrast colors within two planes of cone-contrast space. They presented
chromatic stimuli within the constant S-plane as well as in the constant L-M-plane, in which
18
GENERAL INTRODUCTION
Experimental objectives
either S-cones or differentially combined L- and M-cones have no contribution to the
observed responses. To derive a roughly complete characterization of chromatic response
patterns in human visual cortical areas many more planes in cone-contrast space can be
tested. We use the same two planes and extend the chromatic stimulus array to essentially five
planes in cone-contrast space. Additional planes constitute the constant M-plane, constant
L+M-plane and the constant L-plane, in which, either M-, L+M or L-cones have no
contribution to the measured responses respectively.
Another main objective of this thesis is to compare how chromatic response patterns in
extra-striate areas differ as compared to striate cortex. For that reason, extra-striate regions
hV4, V3A and MT+ are compared to primary visual cortex (V1) with respect to their
responses to the above described chromatic stimulus array. Furthermore it is of interest to
what extent the response characteristics may change as different stimulus velocities and
chromatic contrasts are applied in order to achieve a more complete understanding of cortical
color and motion processing in the participating regions.
A prior study has explored some effects of presenting cone-contrast colors at different
temporal frequencies (Liu & Wandell, 2005). The authors reported cone-contrast sensitivity
curves for three color-directions at two velocities. We are able to extend these findings to
measure cone-contrast sensitivity curves for thirteen directions and two velocities and
additionally measuring five of these directions at three velocities. We are thus able to report
speed tuning curves for five cone-contrast space directions in visual areas V1, hV4, V3A and
MT+.
These extensive measurements allow for strong conclusions regarding chromatic color
contrast preferences throughout visual cortex and its dependence on speed. We are able to
relate chromatic speed tuning properties in several visual areas to psychophysically derived
motion channels (Gegenfurtner & Hawken, 1996). Additionally, the results provide clues for
previously uncertain cortical sources regarding motion detection thresholds using
achromatopsic patients (Cavanagh et al., 1998). Furthermore, mixed color responses will be
compared to stimuli emphasizing cardinal color directions in human visual cortex. This is of
interest because recent neurophysiological studies reported that, starting in striate cortex,
neuronal response preferences may shift away from colors solely stimulating the cardinal
directions as found in monkey LGN (Lennie et al., 1990; Solomon & Lennie, 2005; Conway
& Livingstone, 2006).
The next section will describe in detail how stimulus control was achieved, how conecontrast colors have been computed as well as a short description of the three applied fMRI
analysis methods.
19
GENERAL METHODS
Stimulus control
2 GENERAL METHODS
2.1 Stimulus control
2.1.1 Classifying colors
Colors are often represented in three-dimensional color spaces. The choice of the basis
(axes) differs depending on motivation and assumptions. Some color spaces are related in a
way that allows for a relatively simple mapping between them, in cases when they have
similar theoretical background. This thesis measures and describes physiological correlates of
color, contrast and motion velocity variations. A natural choice for a color space to
parameterize stimuli for such measurements is the physiologically motivated 3D-conecontrast space. This color space incorporates the spectral sensitivity of the three cone classes
in the human retina and the contrast normalization implemented in the visual system.
2.1.1.1 Physiological color spaces
A well known physiologically motivated color space is the DKL-space (Krauskopf et
al., 1982; Derrington et al., 1984). It constitutes a three-dimensional extension to the twodimensional MacLeod and Boynton chromaticity diagram (MacLeod & Boynton, 1979).
DKL-space assumes that two chromatic mechanisms exist in which signals from different
cone-classes are combined (L-M, S-(L+M)). Activity regarding these mechanisms is
represented along two orthogonal axes forming an isoluminant plane. The third axis is
(nearly) orthogonal to the isoluminant plane and represents the luminance mechanism. The
excitation of the three mechanisms by the adapting background field constitutes the origin of
this color space, while excursions along either axis represent excitation through visual stimuli
that selectively stimulate the mechanism represented by that axis. Each axis is considered
cardinal because signals varying along these directions are carried along separately adaptable
pathways (Krauskopf et al., 1982).
The three axes of the cone-excitation space are proportional to the quantal catch of the
three receptor classes (L, M, S). To represent a test stimulus around the adapting field the
quantal catch value from the adapting field is subtracted from the quantal catch values of the
test stimulus for each receptor (axis). It can thus be defined as to how much difference in
quantal catch can be expected from the test stimulus for each receptor class determined by the
cone excitation difference values. The DKL-space is a linear transformation of the axes of
local cone excitation space (Eskew et al., 1999).
20
GENERAL METHODS
Stimulus control
The color space used to compute the color stimuli used in this thesis is the conecontrast space. It has been recently applied in neurophysiological and behavioral experiments
and thus facilitates comparison of results with this work (Engel et al., 1997b; Wade et al.,
2002; Chatterjee & Callaway, 2003; Liu & Wandell, 2005; Solomon & Lennie, 2005; Conway
& Livingstone, 2006). It is based on physiological mechanisms and takes the adaptational
state of the visual system into account. The later is achieved by dividing the cone excitation
difference between adapting background and test stimulation for each cone class, and scaling
it with the adapting value (the cone excitation by the background) for each cone class. Hence
the three axes represent proportional changes for each cone class of the adapting light level
after Webers-Law. In other words: On the three axis we plot ΔS/S, ΔM/M and ΔL/L
respectively, while ΔS, ΔM and ΔL represent the change in each cone class elicited by the test
stimulation as compared to the background excitation (adapting level). L, M and S represent
the excitation of the background for each cone class by which the excitational difference is
then divided. The cone-contrast space represents sensitivity scaling effects as they are found
in the visual system. Cone-contrast coordinates can be transferred into cone excitation space
as long as the adaptational level, the background, is kept constant.
The three cardinal mechanisms (Krauskopf et al., 1982) can be incorporated into conecontrast space. The L-M-mechanism (R/G) would form a diagonal axis within the iso-S-coneexcitation-plane spanned by L- and M-cones, as long as an excitational increase of the Lcone-system is associated with an excitational decrease of the M-cone-system, again, relative
to the background. The two opponent colors that stimulate the L-M mechanisms best are
either red, weighting the L-cones with 1 and the M-cones with -1, and green, weighing the Lcone-system with -1 and the M-cone-system with 1 (Cole et al., 1993; Sankeralli & Mullen,
1996; Newton & Eskew, 2003). Keeping the ratios of the weights equal ensures equal
stimulation of both receptor classes, albeit of opposite sign for one color.
Another important consequence of assigning equal weights is, that within one plane of
the three-dimensional color space, simple sign changes of axes weights generate stimuli along
orthogonal mechanisms. For instance assigning equal weights of equal sign to the respective
L- and M-cone axis (1*L + 1*M or -1* L -1*M) generates stimuli along a proposed
luminance mechanism which also spans a diagonal within the iso-S-cone-excitation-plane
lying orthogonal to the afore described L-M mechanism.
Arranging the cone-contrast color stimulation in this way has the following
advantages. Firstly, stimulation of either cardinal mechanism (L-M, L+M or S) does not
stimulate either mechanism of the other two, a consequence of orthogonality, which is
essential to study mechanism responses throughout cortical regions. Secondly, any additional
21
GENERAL METHODS
Stimulus control
linear combination of the 3 cone classes can easily be implemented to study higher stage
cortical color mechanisms. A pure, perceptually, achromatic B/W-axis can for instance be
introduced, weighting all three cone systems equally (L+M+S; see table), by forming one
main diagonal within this three-dimensional-space. Thirdly, activity can be plotted in terms of
physical stimulus content and thus excitation of the cone systems and not in terms of
detection/discrimination mechanisms the properties of which may change as different
stimulus form or temporal frequencies are employed (Wandell, 1995; Wandell, 1999).
Table 1 describes the weights we have used to form the axes in cone-contrast space
along which cone-contrast stimuli were computed for the experiments reported in this thesis.
The associated color perceptions can be judged in Figure 5.2.
L
M
S
L+M
L-M
L+S
-L+S
M+S
-M+S
L+M+S
L-M+S
L-M-S
S-(L+M)
L
1
0
0
1
1
1
-1
0
0
1
1
1
-1
M
0
1
0
1
-1
0
0
1
-1
1
-1
-1
-1
S
0
0
1
0
0
1
1
1
1
1
1
-1
1
Table 1: Weights for multiple vector directions in cone-contrast-space as they are
employed in the presented experiments to study chromatic properties of human visual cortex.
Throughout this thesis terminology like luminance and isoluminance is employed.
These terms are frequently used in the literature sometimes even without precise definition.
The following sections aim to clarify the origin, definition and methodological background
that led to the establishment of these terms.
2.1.1.2 Luminance
The definition of luminance has been derived from psychophysical measurements on
humans by the Comission Internationale de l’Eclairage (CIE) in 1924 and represents the
relative luminous efficiency of a human standard observer under photopic conditions (CIE,
1926). The function describes how sensitive the human visual system is to lights at every
wavelength of the visible spectrum. The result is the spectral luminosity function, V(λ).
Knowing this standard luminous efficiency function enables one to calculate visual
22
GENERAL METHODS
Stimulus control
effectiveness directly from the spectral distribution of energy in a given light source.
Specifically, the luminance value of a light is calculated as the integral over wavelength of the
radiance of that light weighted with the spectral luminosity function V(λ).
Physiological correlates of this sensitivity curve are ambiguous, making the exact
choice of a cone weight ratio to isolate an alleged pure luminance channel a somewhat
difficult matter. It is believed that at some stage beyond the receptor level, signals from the
three classes of cones are combined to yield three transformed signals. V(λ) may represent the
spectral sensitivity of one of these postreceptoral mechanisms, while the possibility of all
three contributing to V(λ) could not be ruled out to date. Lennie and collegues speculate that
V(λ) may be explained as linear combination of the signals from L- and M-cones,
approximately in proportions as they exist in the human retina (Lennie et al., 1993). The ratio
of L- and M-cones across observers is highly variable (Hagstrom et al., ; Brainard et al., 2000;
Krauskopf, 2000b) rendering such criteria nonspecific. General observations exist, stating that
the V(λ)-curve could be approximately derived by summing the separately measured L- and
M-cone sensitivity curves, but weighting the L-cone sensitivities with a factor of 2 (Nerger &
Cicerone, 1992; Dobkins et al., 2000a), resembling an L:M proportion of 2:1. Krauskopf finds
the retinal L/M ratio to be equal to 1 (Krauskopf, 2000b), while other investigators report
retinal L/M-ratios to lie between 0.82 and 9.71 (Brainard et al., 2000). A relative invariance of
these ratios regarding foveal to peripheral position was reported (Nerger & Cicerone, 1992),
while it seems as if the ratios approach equilibrium in more peripheral positions (Krauskopf,
2000b; Wikler et al., 2004).
A requirement for our visual stimulation was to design targets that emphasize expected
post-receptoral mechanisms while keeping stimulation of other mechanisms minimal
(Derrington et al., 1984). For that reason, the stimulus configuration to excite the luminance
mechanism was chosen to assign equal weights to L- and M-cones, and thus follows the
orthogonality constraint with respect to the R/G-mechanism, taking the difference of the
equally weighted L- and M-cone signals. This choice also provides orthogonality with respect
to the B/Y mechanism varying the signs of the equally weighted S-cone signals.
While there is wide agreement that S-cones have negligible contribution to V(λ)
(Wyszecki & Stiles, 1982; Lennie et al., 1993), it is worth noting that most techniques derive
V(λ) with spatial or temporal frequencies that discriminate against signals from short wave
sensitive receptors (Lee et al., 1988; Lennie et al., 1993). In addition the form of the curve
may change as different techniques are applied to derive it (Lennie et al., 1993). The method
that formed the basis for the CIE 1924 standard is the heterochromatic flicker photometry
(HFP), in which an observer adjusts the intensity of two flickering lights until the perceived
23
GENERAL METHODS
Stimulus control
flicker is minimized. At the point of minimum flicker the two lights are believed to be
equiluminous or isoluminant.
2.1.1.3 Isoluminance
It has been reported that chromatic and luminance motion mechanisms appear to have
fundamentally different properties (Ramachandran & Gregory, 1978). As indicated above, the
methodological dependence of luminance measures directly applies to isoluminance
measurements accordingly. It is thus of interest in what way these two entities relate and how
they are typically derived. Ex ante, the definition of isoluminance is tightly coupled to the
methodology by which it is derived.
Heterochromatic flicker photometry (HFP) uses fairly high flicker rates (~15 Hz) to
equalize luminosity intensities of two colors with different spectral compositions and thus
derives at some magnocellularly biased luminance match (Lee et al., 1988; Lennie et al.,
1993; Dobkins et al., 2000; Dobkins et al., 2000a), which can be modeled as the sum of Land M-cone sensitivities (L+M) (but see section 2.1.1.2). There is reason to believe that
differential L- and M-cone signals (L-M) might also contribute to luminance matches as
temporal frequency is lowered (Cavanagh et al., 1987 ; Webster & Mollon, 1997; Dobkins et
al., 2000; Dobkins et al., 2000a). The method of heterochromatic brightness matching (HBM)
is thought to rely on signals from both L+M and L-M mechanisms and involves directly
assessing and matching the brightness of two stationary colors (Dobkins et al., 2000). The two
colors will be perceived as equally luminous, and thus isoluminant, when the signals of L- and
M-cone excitation, produced by the spectral composition of one color, equal the L- and Mcone excitation produced by the spectral composition of the other color. As expected from a
task which theoretically employs two mechanisms (HBM), the two colors set to be equally
bright in HBM are mostly not perceived equiluminant in the HFP-task, theoretically relying
only on one mechanisms (Webster & Mollon, 1993; Smith et al., 1994). Webster and Mollon
inferred that at low temporal frequencies both perceived lightness and minimum-motion
settings appear to depend on channels that do not represent luminance and color
independently (Webster & Mollon, 1993).
The arguments above make clear, that concepts, such as luminance or isoluminance
are highly dependent on the methodology by which they are derived. The main interest of this
work lies in cortical responses to color, for which a physically motivated stimulus control
seems more appropriate as a perceptual but physiologically biased one. We thus indicate our
colors as chromatic stimuli as opposed to isoluminant, while the luminance stimulation
follows conventional nomenclature (but see next section). Stimuli used in this thesis follow
24
GENERAL METHODS
Stimulus control
the conventions of the cone-contrast space and its applications (see above), because the
spectral cone sensitivities have a solid, task independent, foundation. Using cone-contrast
space, the stimuli can be parameterized, which would not be possible using just isoluminance
criteria (Schnapf et al., 1987; Stockman & Sharpe, 1998; Stockman et al., 1999; Stockman &
Sharpe, 2000).
2.1.1.4 Luminance vs. achromatic stimuli
Luminance and perceptually achromatic stimuli are often erroneously employed as
being synonymous. Perceptually these two stimulation types can be very easily distinguished.
Neuronal projections from the retina to LGN constitute at least three independent
mechanisms (Derrington et al., 1984). A first one responds to the difference of L- and Mcones (L-M) and a second responds to the difference between the excitation of the S-cones on
the one hand and the sum of L- and M- cones on the other hand (S-(L+M)). These two
mechanisms are chromatic mechanisms because stimuli that isolate the R/G mechanism best
(L-M) appear red/green and stimuli that excite the B/Y mechanism best (S-(L+M)) appear
blue/yellow. In order to isolate the B/Y mechanism best however, stimuli should only vary the
modulation of the S-cones to control for luminance contribution. These stimuli appear
greenish/purple.
A third axis, perpendicular to the chromatic axis is believed to represent the luminance
axis. As indicated above, a luminance mechanism may sum L-and M-cone responses and omit
S-cone signals (Judd 1951, Wyszecki & Stiles 1982). Stimuli exciting only the L- and Mcones with the same sign appear purple/yellow and not black and white (achromatic). A
stimulus designed to excite all three cone classes equally, however, would appear black and
white and can thus be referred to as perceptually achromatic. This axis, unfortunately, is not
precisely orthogonal to the S and L-M axis in cone-contrast space. A perceptually achromatic
stimulus would thus also activate the presumed luminance mechanism (L+M) and the B/Y
mechanism (S). We stress this, because we will show that cortical responses to L+Mcompared to L+M+S-stimuli differ substantially especially at high speeds. Unless indicated
differently, we refer to L+M+S-modulation as achromatic, L+M-modulation as luminance and
opponent (difference in sign) modulation as chromatic ((L-M & -L+M), (S & -S), (S-L-M & S+L+M)).
2.1.1.5 Color vs. luminance
Experiment 2 will investigate in what way responses to chromatic stimuli in several
visual areas differ as the targets are moved at different velocities. In order to allow for
25
GENERAL METHODS
Stimulus control
equitable comparison between responses in visual areas to the various color- and the
luminance-stimulation and clearly assess the effects of a given additional stimulus attribute, in
the case of experiment 2, pattern velocity, it is crucial to match all stimuli regarding their
cone-contrast excitation. It is well established that low-level visual functions depend heavily
on the magnitude of the input signal (contrast). Nearly all neurons in the visual system give
faster and stronger responses to stimuli of higher contrasts (Albrecht & Hamilton, 1982). This
effect can also be observed using fMRI in human V1 (Boynton et al., 1996) and extra-striate
areas investigated with incremental chromatic-contrast (Engel et al., 1997b; Wandell et al.,
1999; Liu & Wandell, 2005). In order to measure response differences, arising from
differential chromatic preferences in the visual areas, independent of the applied (chromatic)
contrast, we compute all stimulus colors along cone-contrast space direction in a way, that
their contrast modulation matches physically at the initial retinal input stage (Wandell, 1995;
Kaiser & Boynton, 1996). It becomes thus possible to measure sensitivity hierarchies for the
visual areas, derived solely from chromaticity differences as one compares responses at the
same pattern velocity. Additionally, it is possible to report response differences depending on
pattern velocity as responses to the same color are compared, moving at different speeds. One
can thus be certain that observed response differences did not arise from contrast differences
but must be associated to the tested stimulus attributes as for instance various colors or
speeds.
The following section describes how cone-contrast of the stimulation is controlled.
2.1.2 Computing cone-contrast colors
2.1.2.1 Light to cone excitations
Light sources in color experiments are typically projectors or monitors, which generate
light through three red (R), green (G) and blue (B) phosphors (guns) at each pixel. Each single
phosphor class emits light over the entire visual spectrum; radiated energy however is not
equal at all wavelengths. For instance, the R-gun emits more energy in the long-wave range of
the visual spectrum, where light appears red. The G-gun emits more energy in the middlewave range of the visual spectrum, where light appears green and the B-gun emits more
energy in the short-wave range of the visual spectrum, where light appears blue.
Two sets of data are essentially required in order to compute the intensity levels by
which each class of retinal receptors is stimulated by a given light. Firstly, the spectral
sensitivity of each cone class and secondly, the energy distribution (spectral density) of the
given light over the whole visible spectrum. The spectral sensitivity of each cone class is
26
GENERAL METHODS
Stimulus control
provided in the literature (Schnapf et al., 1987), and tables can also be downloaded from the
internet (www.cvision.ucsd.edu). The spectral density of the light source can be measured by
the means of a photospectroradiometer for a set of wavelengths in the visible part of the
spectrum.
As stated above, each cone class is sensitive to light of all wavelengths by a certain
amount, but it is most receptive within a certain wavelength range. If one wishes to compute
how strongly a cone class, let’s say the L-cones (l), is excited at one particular wavelength (λ),
one must weight the spectral sensitivity of the L-cone class at that particular wavelength (l(λ))
with the emitted energy from the light source, e.g. R-gun at that wavelength (r(λ)). In order to
compute the excitation of one cone class, e.g. the L-cone system, by the, e.g. R-gun of a CRT
over all wavelengths, the spectral density of the R-gun must be weighted by the spectral
sensitivity of the L-cone system. The products of the excitation values at every wavelength
step give the excitation in the L-cones, elicited solely by the R-gun. This can be expressed as:
LR = ∑ l (λ ) * r (λ ) .
λ
LR = L-cone excitation values by R-gun,
l(λ) = L-cone sensitivity at wavelength λ,
r(λ) = spectral density (energy radiance) of the R-gun at every wavelength of the visual
spectrum.
To compute the excitation of the L-cones elicited by all three phosphors, each spectral
density function per phosphor has to be weighted with the L-cone system sensitivity function
separately and then summed:
L = ∑ l (λ )* r (λ ) + ∑ l (λ )* g (λ ) + ∑ l (λ )* b(λ ) .
λ
λ
λ
L
= L-cone excitation,
l(λ)
= L-cone sensitivity at wavelength λ,
r(λ),g(λ),b(λ) = spectral density (energy radiance) of the R-, B-, and G-gun at wavelength λ.
27
GENERAL METHODS
Stimulus control
The excitations elicited by all three phosphors in all three cone-types can be calculated
by the following system of equations:
L = ∑ l (λ ) * r (λ ) + ∑ l (λ ) * g (λ ) + ∑ l (λ ) * b(λ ) ,
λ
λ
λ
M = ∑ m(λ ) * r (λ ) + ∑ m(λ ) * g (λ ) + ∑ m(λ ) * b(λ ) ,
λ
λ
λ
S = ∑ s (λ ) * r (λ ) + ∑ s (λ ) * g (λ ) + ∑ s (λ ) * b(λ ) .
λ
λ
λ
L, M, S
= cone excitation values
l(λ),m(λ),s(λ) = cone sensitivity values at wavelength λ
r(λ),g(λ),b(λ) = spectral density (energy radiance) of the R-, B-, and G-gun at wavelength λ
All of the above is valid, at any one specific radiance intensity of the RGB-guns. The
intensity level for each gun can be altered separately. For this reason an additional vector is
introduced, representing the intensity levels for every single gun. The above cone excitation
computation can be summarized as a 3x3 transformation matrix. This transformation matrix
can now be multiplied with the intensity vector of each single gun depending on the adjusted
intensity of each gun on the CRT-screen:
⎡
l (λ ) * r (λ )
⎡L⎤ ⎢∑
λ
⎢
⎢M ⎥ =
m(λ ) * r (λ )
⎢ ⎥ ⎢∑
λ
⎢
⎢⎣ S ⎥⎦
⎢ ∑ s(λ ) * r (λ )
⎣ λ
⎤
∑λ l (λ ) * g (λ ) ∑λ l (λ ) * b(λ ) ⎥ ⎡ R ⎤
⎥
∑λ m(λ ) * g (λ ) ∑λ m(λ ) * b(λ )⎥ ⎢⎢G ⎥⎥
⎥⎢ ⎥
∑λ s(λ ) * g (λ ) ∑λ s(λ ) * b(λ ) ⎥⎦ ⎣ B ⎦
.
R, G, B = Scaling values for Red-, Green- and Blue-gun (between 0 and 1)
Values of the intensity vector typically range between 0 and 1, where 0 means that the
gun does not emit any photons, while 1 represents the maximal amount of photon emission
possible. In conventional monitors the increase of emission of each single gun is rarely a
linear function of the RGB-values used in the intensity vector. This is however assumed when
using the above matrix multiplication. It is thus essential to measure the intensity function for
each gun at each intensity level (typically 0 to 255) and linearize the intensity curve. This is
achieved by creating lookup tables in which for instance a doubling of intensity is actually
accompanied with a doubling of photon emission. This point is stressed because it entails
physiological reasoning. Throughout this thesis, cone-contrast will be represented as the
vector length of the three dimensional vector s, representing the cone-contrast in the direction
of the vector in cone-contrast space.
28
GENERAL METHODS
Stimulus control
s = ( ΔL/L, ΔM/M, ΔS/S )
(Three dimensional color vector).
|s| = [ Δ(L/L)² + Δ(M/M)² + Δ(S/S)² ]1/2
(RMS-cone-contrast)
L, M, S
= excitation by the background for each cone class,
ΔL, ΔM, ΔS
= excitation by the test color in each cone class,
ΔL/L, ΔM/M, ΔS/S = change in each cone class elicited by the test stimulation, as compared to
the background excitation (adapting level).
In the experiments presented here, RGB-values had to be computed to produce a
certain cone-contrast. This approach will be described in the next section.
2.1.2.2 Cone excitations to light
In order to derive RGB-values which stimulate the visual system at a desired conecontrast, the above equation must adopt the appropriate values for L, M and S (e.g.: [L+L*0.1,
M+M*0, S+S*0 ]). This cone-contrast for instance would stimulate the L-cone system 10%
more than the background, while keeping stimulation of the other two cone classes equal. In
order to obtain the RGB-values that produce this contrast the inverse of the transformation
matrix must be multiplied with the LMS-vector.
The afore mentioned device dependence on spectral and intensity distributions for
each gun poses limitations regarding maximal attainable cone-contrasts. Along some conecontrast space directions, for instance L-M, the highest maximal cone-contrast is lower than
any other direction, because the M- and L-cones have highly overlapping sensitivity
functions, which means that the probability of catching photons is highly similar. It follows
that strongest limitations occur when two cone types need to be stimulated differentially, as
necessary for L-M stimuli. Figure 2.1 illustrates two examples:
29
GENERAL METHODS
Stimulus control
A
B
S
ML
S
ML
Figure 2.1: Stimulus control for cone-contrast stimulation
A: L-M-cone-contrast stimulus control: Right panel: 14% L-M RMS-cone-contrast radial sinusoidal
grating (annulus) as presented in Experiment 2. Left panel: Spectral energy plot of the stimulus
components as measured with a pr650 spectroradiometer. Solid lines (spectra of the components):
gray (background), pink (L-M), green (-L+M). Dotted lines (normalized cone sensitivity functions): blue
(S-cones), green (M-cones), red (L-cones), vertical black dotted line (lights with wavelengths beyond
this point excite the L-cones more than the M-cones). Note: averaging the spectral radiance over
wavelength for both test colors would result in the background spectral radiance. A L-M stimulus
contains higher spectral radiance in wavelengths exciting the L-cones than the background resulting in
a reddish percept. A –L+M stimulus contains more spectral radiance in wavelengths exciting the Mcones than the background resulting in a greenish percept. A higher spectral energy exciting L-cones
in the L-M color is counterbalanced with a lower excitation of the M-cones and vice versa, for the –
L+M color. The color spectra do not diverge from the background spectra for wavelengths, which
would excite the S-cones. This stimulus configuration isolates the L-M (R/G) mechanism.
30
GENERAL METHODS
Magnetic resonance imaging
B: S-cone-contrast stimulus control: Right panel: 14% pure S- RMS-cone-contrast radial sinusoidal
grating (annulus) as presented in Experiment 2. Left panel: Spectral energy plot of the stimulus
components as measured with a pr650 spectroradiometer. Solid lines (spectra of the components):
gray (background), blue/purple (+S), greenish/yellow (-S). The dotted lines represent the normalized
cone sensitivity functions as in A. Note: averaging the spectral radiance over wavelength for both test
colors would result in the background spectral radiance. A +S-color contains higher spectral radiance
in wavelengths exciting the S-cones than the background resulting in a bluish/purple percept. A –Scolor contains less energy in wavelengths exciting the S-cones than the background resulting in a
greenish/yellow percept. The color spectra do not diverge from the background spectra for
wavelengths, which would excite the L- or M-cones. This stimulus configuration isolates the S-cones.
2.2 Magnetic resonance imaging
Functional magnetic resonance imaging must be considered the fastest growing
application for noninvasive neurophysiological methods for in vivo human brain imaging. The
technique is based on metabolic processes leading to changes of oxygen content and blood
flow in the living brain. These factors are coupled with neural activation, albeit indirectly.
High spatial resolution is an advantage this method provides. Experiments in this thesis put
emphasis on spatial precision to distinguish responses in different areas within human cortex,
making fMRI the method of choice.
Brain activity is associated with depolarization of membrane potentials of neurons in
active regions. Neural activity is energy consuming, at the expense of oxygen. Within a few
seconds of regional neural activity, compensatory processes lead to local increase of blood
flow and blood volume. This process is known as neurovascular coupling. The local increase
of oxygenated blood exceeds the metabolically needed oxygen consumption, which leads to a
netto increase of oxygenated hemoglobin in the vessels and capillaries of active brain tissue.
This local increase of blood oxygenation can be detected by the blood oxygenation level
dependent (BOLD) functional magnetic resonance imaging (fMRI) technique, using for
instance an echo-planar imaging sequence (EPI), exploiting the different magnetic properties
of oxygenated and deoxygenated blood (Thulborn et al., 1982). While oxygenated
hemoglobin shows diamagnetic properties (magnetically transparent), deoxygenated blood is
paramagnetic (not magnetically transparent). The ratio between dia- and paramagnetic
hemoglobin is accountable for the MR-signal, which can be observed in contrast changes in
the MR-images, where active cortical regions are typically brighter than inactive regions. This
spawned the term: BOLD-contrast or BOLD-signal (Blood Oxygenation Level Dependent)
(Ogawa et al., 1990; Turner et al., 1991; Ogawa et al., 1992).
The differences of signal strength between stimulus condition and rest condition
depends on the static magnetic field strength of the MR-magnet and is typically less than 4%
for scanners up to 3T (Tesla). The low signal to noise ratio may be accommodated by multiple
31
GENERAL METHODS
Magnetic resonance imaging
presentations of the experimental stimulation and additional averaging of the stimulus related
images.
Within this thesis, three different fMRI-analysis methods are employed. Common to
all three methods is an initial motion correction of the acquired brain volumes removing
movement artifacts of the subject’s head during the MR-measurements. The artifacts are
caused by the subjects’ motion, respiration and pulse. The rigid-body-motion-correction
algorithm implemented in the SPM99 software was used (SPM99, Wellcome Department for
Cognitive Neurology, London, UK). All three methods initially reconstruct the fMRI time
series in each voxel independent of neighboring voxels, which entails linear trend removal
from the individual voxel time series.
2.2.1 Method 1 – phase encoding
This method was used for retinotopic mapping of the borders of the visual areas and in
MT localizer experiments. A prerequisite for this method constitutes a periodic presentation of
the visual stimuli. A simple example for only two stimulus conditions will illustrate this
method.
Any cortical region that represents the entire visual field retinotopically should by
definition contain neurons that respond to visual stimulation in the center of the visual field
and spatially separated neurons that respond to visual stimulation outside of the central visual
fields. In order to identify these regions a simple periodic visual stimulation may be used that
alternates between a center and a surround stimulus at constant frequency. We map these
regions routinely by presenting a center stimulus for 12 s followed by a surround stimulus for
12 s, while sampling brain activity every 2 s. This stimulus sequence is presented 20 times per
run leaving us with an alternating frequency of 20 cycles per run which lasts for 8 minutes
equivalent to 240 acquired scans or brain volumes.
Figure 2.2 illustrates the stimuli, stimulation time course and corresponding BOLD
response time courses for small regions of interest in primary visual cortex. Dotted red and
blue boxcar functions describe the time course of the center and surround stimulus onsets
respectively. They are out of phase by 180°. Firstly, the fMRI time series for each voxel is
generated by concatenating the intensity values at every single voxels’ coordinates from the
scanned brain volumes measured every 2 s. Secondly, the time series for each voxel is
computed as a percentage modulation of its mean intensity value. Activity elicited by the
stimulation is measured by fitting the phase and amplitude with a sinusoidal function (black
dotted line in Figure 2.2) at the stimulus alternation frequency (1/20 cycles/scan = 20
presentations) to the time series, essentially by performing a Fourier analysis. At each voxel
32
GENERAL METHODS
Magnetic resonance imaging
the response phase indicates the preferred stimulus. Figure 2.2 demonstrates how voxels,
taken from separate locations in V1, follow either stimulus time course. Note how the BOLDresponse time courses differ regarding their phase while the frequency is equal and
corresponds to the stimulus alternation frequency. Red arrows indicate different phases
following the preferred visual stimulation within the underlying voxels from which these two
responses were taken (ROIs).
Center stimulus response time course
A
Time [sec]
Center stimulation
D
F
B
V3d
V2d
Surround stimulation
G
V1
V2v
V3v
E
C
Surround stimulus response time course
Time [sec]
Figure 2.2: Phase Encoding. A and C: The BOLD-activation time courses elicited by a
center-surround stimulus in occipital cortex. The solid orange line represents the measured response
time course for a region of interest (ROI) indicated as a blue circle in B. The dotted red line indicates
the stimulus on- and offset for center stimulus presentation as illustrated in panel D. B: Center
(orange) and surround (blue) activation on flat map representation for one hemisphere. Solid lines
represent borders of visual areas (V1=blue, V2=red, V3=yellow). C: The solid blue line represents the
measured response time course for a region of interest (ROI) indicated as yellow circle in B. Dotted
blue line indicates the stimulus on- and offset for surround stimulus presentation as illustrated in panel
E. D: Center stimulus. E: Surround stimulus. F: Red dotted line shows stimulus on- and offset for
center stimulus presentation. Black dotted line represents the alternation frequency for the center
stimulus. G: Blue dotted line stimulus on- and offset for surround stimulus presentation. Black dotted
line represents the alternation frequency for the surround stimulus. Red arrows in A and F show the
short phase of the alternation frequency for center stimulation. Red arrows in C and G exemplify the
longer phase of the alternation frequency for surround stimulation.
33
GENERAL METHODS
Magnetic resonance imaging
Many more stimulus conditions and orders can be presented as long as the alternation
frequency is kept equal and constant. This will be demonstrated in Experiment 1. This
procedure implicitly models the BOLD-response as a harmonic function (Bandettini et al.,
1993; Press et al., 2001), which differs from Method 2.
2.2.2 Method 2 – canonical BOLD fitting (GLM)
The basic principle of method 2 constitutes a comparison of a theoretically predicted
time course with the measured BOLD time series at every voxel. The background for the
theoretical time course is the characteristic BOLD-signal distribution over time known as the
hemodynamic response function (HRF) (Figure 2.3). Specifically the assumption, that with
the onset of neural activity elicited by some experimental stimulation (stimulation condition),
the HRF raises with temporal delay in an active cortical region and drops with the absence of
stimulation as neural activity decreases accordingly (rest condition) (see Figure 2.4 and 2.5).
Figure 2.3: A canonical BOLD-response function as it is used in SPM
The BOLD-response onset lags behind stimulus onset, hypothetically given at 0 seconds, a wellknown feature of the BOLD contrast. The maximum BOLD-response amplitude is reached 4-10 s after
stimulus onset while the BOLD relaxation period after stimulus offset can range from 10-20 s.
The comparison of the empirically measured and the theoretically predicted time
courses for every voxel can be accomplished by application of multiple regression analysis
using matrix notation and the general linear model (GLM). The expected BOLD-response
time courses are coded in a design matrix, in which each column represents an experimental
condition and each row one image acquisition of the entire fMRI measurement, depicted in
Figure 2.4. For each condition, the theoretically predicted response time course (predictor
variable) is obtained by convolving the stimulus time course for each condition with the same
34
GENERAL METHODS
Magnetic resonance imaging
assumed canonical HRF-function, as depicted in Figure 2.3. Hence the design matrix should
explain the variance of the measured empirical signal time course with minimal residual error.
The goal is to minimize the least-square-error by assigning appropriate weights to the
predictor variables using multiple regression and the GLM.
In common analysis tools such as SPM99 (http://www.fil.ion.ucl.ac.uk/spm) multiple
regression is applied at every voxel. Within each voxel, the value of the weight at a predictor
variable represents the contribution to the measured time course for that condition. Cortical
regions are identified which follow a predicted time course for a given condition and thus may
be associated with the cortical processing of that condition. Figure 2.4 illustrates the GLM
weighting procedure for one voxel using the center/surround stimulation as an example.
Interpretatively, one would conclude that the voxel from which the BOLD response time
course in Figure 2.4 was taken responds stronger and follows the expected time course of the
first condition substantially better. Hence neural populations within that voxel represent the
central visual field. This procedure is applied to every voxel resulting in maps of brain activity
representing strength of BOLD-response and fitting values (betas) for any predefined time
course and condition.
In order to find cortical regions that prefer one stimulus condition over the other, as in
method 1, respective statistical maps can be calculated using the subtraction method. A
response map elicited by, for instance condition one, is subtracted from a response map
elicited by condition two, resulting in a difference map which contains only voxels, exceeding
a predefined significance level. These statistically significant voxels regarding the difference
between the two conditions, yielded BOLD-response time courses, which fit the predicted
response of condition two significantly better than condition one. Since many voxels are
compared simultaneously, the ‘problem of multiple testing’ arises which entails that at
constant alpha of 5% (α= first order error) the probability of observing false positive results
increases proportionally to the number of conducted single tests (number of compared
voxels). To control for this problem many procedures exist, but the most prevalently used is
the Bonferroni correction, which corrects the applied significance level α by the number k of
conducted independent tests (αcorrected = α/k). Another approach to decrease the number of
independent tests is by restricting the analysis to one or a few pre-defined regions of interest
(ROI-analysis) the number of comparisons by ROI-analysis. A priori hypothesis are necessary
to define ROIs, amongst which are for instance anatomical landmarks or localizing regions of
interest based on their functional properties.
35
GENERAL METHODS
Magnetic resonance imaging
Figure 2.4: GLM weighting procedure for one voxel using the center/surround
stimulation.
Blue lines plot the measured BOLD response values in that voxel. The dark gray lines represent the
predicted time course for the first condition (center stimulation) in the upper panel and second
condition (peripheral stimulation) in the lower panel respectively. The red lines indicate the fitted
response function generated by SPM99. In the upper panel the fitted response follows the predicted
time course for the first condition, while in the lower panel a rather poor fit between the predicted time
course of the second condition and the measured values is obvious. The GLM would hence attribute a
high weight (beta) to the explanatory variable condition one, while the second explanatory variable,
condition two, would be weighted substantially lower and can also become negative. The two different
time courses can statistically be compared by means of F-tests or t-tests in which the absolute weight
(beta) difference determines the effect size.
2.2.3 Method 3 – deconvoluting BOLD-response functions
The following section is divided into two parts. The first part will describe the
assumptions underlying a reliable deconvolution of time-series data by contrasting block- and
event-related designs. The second part will describe the weighted maximum likelihood
estimate analysis (WMLE) employed in Experiment 3 (Hinrichs et al., 2000).
2.2.3.1 Block design vs. event related design
The major challenge facing every complex fMRI design is the overlapping of BOLD
responses, which occurs when two stimuli are presented closely in time 2 . The overlap effect is
however specifically desired in block stimulation in which, for an elongated period of time,
2
In the case of fMRI, close means within 30s.
36
GENERAL METHODS
Magnetic resonance imaging
the same condition is presented in order to increase the signal to noise ratio. A linear
summation of the BOLD-responses is assumed.
Figure 2.5 illustrates the phenomenon of linear summation considering light flashes as
stimulus events which, presented in succession, compose a block presentation. In panel 2.5A,
a light is flashed for a short amount of time in the beginning of the scanning session. Panel
2.5B depicts the sluggish BOLD response that can be encountered in regions of visual cortex
being responsive to this kind of stimulation. Note, that approximately 4-10 s (2-5 scans,
TR=2), are needed for the BOLD response to reach a maximum. The response curve in panel
2.5B was modeled by convolving the stimulus time course with the canonical HRF depicted
in Figure 2.3. Panel 2.5C shows another light stimulus that was flashed later within the
scanning session and the result is a likewise delayed BOLD response. When both flashes are
presented fairly close in time to one another, as in panel 2.5E, both BOLD response functions
show substantial overlap (2.5F). The resultant BOLD response function is characterized by
increased amplitude, which again was modeled as a convolution of the sum of the single
stimulation time courses with the canonical BOLD function in Figure 2.3.
A
E
B
F
C
G
D
H
Figure 2.5: Linear summation assumption of BOLD-responses.
A: A light flash presented at the beginning of the scanning session. B: Expected BOLD-response for
voxels associated with visual processing from stimulation in A. The response time course represents a
convolution of the stimulus presentation time course in A and the canonical BOLD-response function
(Figure 2.3). C: A light-flash presented 10 seconds after the beginning of the scanning session. D:
Expected BOLD-response from stimulation in C. The response time course is a convolution of the
stimulus presentation time course in C and the canonical BOLD-response function (Figure 2.3). E: Two
light flashes are presented in closer succession. F: Substantial overlap of BOLD-responses generated
by stimulus presentation in E. G: Successive presentation of light flashes forming a stimulus block. H:
Substantial overlap generated by stimuli in G. Additive overlap raises the signal to noise ratio
considerably. The linearity assumption however is not infinitely valid because BOLD-response
amplitudes typically reach a saturation plateau which cannot be exceeded (Boynton et al., 1996).
37
GENERAL METHODS
Magnetic resonance imaging
The desired increasing signal to noise effect of block stimulations becomes clear in panel
2.5G and 2.5H. As light flashes are presented in temporally close succession, the single events
form a block of stimulation (2.5G). The overlapping response functions are assumed to add up
linearly, increasing the signal to noise ratio substantially (Boynton et al., 1996). Nonetheless,
block design stimuli take time, not only because the actual stimulus presentation is rather
long, but also because stimulations containing no experimental variable are typically
presented (baseline condition). The presentation of a baseline condition after the experimental
block-condition allows the large BOLD-response in active regions to relax and provides the
experimenter with a neutral condition, with which the experimental conditions can be
statistically compared.
If a block design is desired in combination with a multitude of conditions however, it
has been shown that a balanced randomization of conditions permits omission of a designated
baseline between the experimental stimulus presentations (Engel, 2005). A balanced
randomization of the stimulus presentation order is subject to the constraint that each
condition is to be preceded by each of the other conditions equally often. This procedure leads
to the same prior block history for every condition and was used to generate the stimulus
presentation sequence for experiment 2. The underlying assumption is, that equal overlap of
BOLD-response functions does not alter response differences between conditions, which may
subsequently statistically be analyzed.
In contrast to the traditional block design fMRI experiment, event related fMRI
experiments have become increasingly popular, addressing research questions, which cannot
be tested using blocked stimulation. Block design disadvantages such as adaptational,
sequential and habituational effects, can be minimized by event related fMRI experiments,
applying randomization and rapid presentation of various experimental conditions (Buckner et
al., 1996; Dale & Buckner, 1997; Josephs et al., 1997; Rosen et al., 1998; Dale, 1999).
Stimuli may be presented with ISIs in the range of a few seconds, taking into account that
subsequent HRFs will overlap significantly (Boynton et al., 1996). Figure 2.5E to 2.5H
showed how overlapping response functions elicited by different types of light flashes sum to
generate the measured BOLD-signal, but implicitly assumed that different light flashes elicit
the same BOLD-response amplitude, which must not necessarily be the case. Figure 2.6
illustrates how the measured BOLD-signal time course can have numerous underlying
BOLD-response functions presumably elicited by different stimulation conditions. However,
responses to the same stimulus may be present in several regions of the brain but shape and
amplitude may vary due to different vascularization and hemodynamic properties of the
38
GENERAL METHODS
Magnetic resonance imaging
constituting cell conglomerates within the respective regions (Logothetis, 2003). This must be
taken into account as rapid event related stimulus sequences are analyzed.
As stated above and exemplified in Figure 2.5, an expected HRF could be modeled by
convolution of the known stimulus time course and a theoretically assumed BOLD-response
function (Figure 2.3). This process can also be reversed, applying a deconvolution of the
measured BOLD-response time course and the known stimulus presentation sequence. First
ideas towards deconvolving fMRI-time series have been proposed by Josephs et al. (1997)
and refined by Dale and Buckner (1997) and Rosen et al. (1998) (Dale & Buckner, 1997;
Josephs et al., 1997; Rosen et al., 1998).
Figure 2.6: BOLD-response functions may differ in shape and amplitude
Shape and amplitude of BOLD-response functions may differ depending on stimulus type as well as
differing hemodynamic response properties of cortical regions
2.2.3.2 Weighted Maximum Likelihood Estimate (WMLE)
Because event-related designs are more susceptible to the shape of the canonical
BOLD-response function, a deconvolution method is chosen to analyze the event-related data
set of experiment 3. The method of choice was the weighted maximum likelihood estimate
(WMLE), which was put forth by Hinrichs et al. (2000), as a refinement of Dale and Buckners
approach (Dale & Buckner, 1997; Hinrichs et al., 2000). While method 1 (phase-encoding)
correlates a predefined sinusoidal function representing the stimulus alternation and method 2
(canonical BOLD fitting) fits a predefined canonical BOLD-response function at the stimulus
presentation time-course, the WMLE approach is a data based procedure with no fixed prior
canonical response function. In short, the procedure simultaneously estimates and weights the
shape of the most probable hemodynamic response function based on the measured BOLDresponse function per voxel, and the presented stimulus event time course. The estimated
39
GENERAL METHODS
Magnetic resonance imaging
weights for each presented condition can then be statistically compared and used to model
cortical sensitivity depending on different levels of each applied condition.
Stimulus presentation in Experiment 3 for instance constituted several color
conditions, which were parameterized in four cone-contrast-stages. The analysis of
Experiment 3 using WMLE yields the most likely HRF, derived from deconvoluting
responses to all color conditions. The resultant HRF function is restricted in its shape and only
the amplitude is free to vary as the fixed HRF is weighted for all cone-contrast-stage
conditions to maximally reconstruct the measured BOLD-response time course at every voxel.
These weights are then averaged across measurements, conducted over several days and can
be used to construct contrast response functions. These contrast response functions can teach
us, how distinct visual areas in the human brain differ, regarding contrast sensitivity
depending on color and speed.
In order to allow for inferences about distinct visual areas, they firstly have to be
precisely located, separately for each subject. If suitable, the respective location can be
confirmed using a functional localizer. These measurements will be presented in Experiment
1. Experiment 2 will investigate speed dependent differences at equal cone-contrasts across
human visual areas and Experiment 3 will extend the speed dependence explorations into the
contrast domain.
40
EXPERIMENT 1- Retinotopic mapping
Introduction
3 EXPERIMENT 1- Retinotopic mapping
3.1 Introduction
Measurements in this thesis regarding motion, color and contrast dependence were
preceded by precise localization of the borders of the visual areas V1, V2, V3 V3A, hV4 and
MT+. Two MR-scanners with different magnetic field strength were used in experiment 2 and
3 respectively. Retinotopical mapping as well as localization of motion sensitive visual area
MT+ were conducted using both scanners, providing an opportunity to confirm spatial
organization of cortical visual field maps for subjects that underwent retinotopical
measurements in both scanners. This chapter reports precise visual field map representation in
human visual cortex and the spatial relation between those regions. The detailed maps will be
used in experiments 2 and 3 to define regions of interest for which investigations of motion,
color and contrast dependencies on BOLD-responses will be conducted.
Figure 3.1: Anatomical landmarks
of visual cortex
PO-Sul
A ventral medial view onto the left
hemisphere of subject ei00. Red star:
occipital pole (posterior lateral end of the
calcarine sulcus). Dotted lines: Red:
Calcarine Sulcus. Green: Fusiform gyrus.
Yellow: Collateral sulcus. Blue: Lingual gyrus.
Brown: Parietal occipital sulcus.
Cal-Sul
Li-Gy
Col-Sul
Fus-Gy
41
EXPERIMENT 1- Retinotopic mapping
Introduction
3.1.1 Retinotopic mapping
The spatial structure of the visual scene falling onto the retina is preserved in LGN and
several cortical visual areas (see Section 1.2.2.3). More precisely, adjacent neurons in the
LGN and visual cortex respond to adjacent points in the visual field. Hence cortical neurons,
spatially organized to represent the visual field, should represent two orthogonal dimensions
of the visual field, namely polar angle and eccentricity. If several spatially distinct cortical
responses to the same portion of the visual field are found, they must correspond to separate
visual field map representations.
A variety of procedures exist that can be used to derive a general idea of cortical
retinotopic organization. It is for instance possible to stimulate solely the vertical or horizontal
meridian of the visual field and apply differential imaging techniques. By doing so, responses
to vertical meridian stimulation will be subtracted from responses to horizontal meridian
stimulation to yield cortical horizontal meridian representation or vice versa to yield vertical
meridian representation (Grill-Spector & Malach, 2001; Hasson et al., 2002; Fize et al.,
2003). This may yield rather imprecise borders for early visual areas (V1-3), while borders of
higher order areas, such as V4, V3A and MT+, elude this procedure. The same is true for
eccentricity measurements. Simply comparing two eccentricities, namely central versus
peripheral, can evidently only yield a rough approximation of the cortical eccentricity
representation (Hadjikhani et al., 1998; Levy et al., 2001; Hasson et al., 2003).
3.1.1.1 Eccentricity
Stimuli measuring the polar angle representation more precisely are preferably several
wedges of high contrast reversing checkerboard patterns, rotating around a central marker on
which the subject fixates. The wedge stimulation is presented at predefined visual field
locations and not only at the horizontal or vertical meridians. The more locations are used and
the thinner the wedge, the more precise the measurement, while restrictions are imposed by
the spatial resolution of the functional BOLD response measurement (Engel et al., 1997a).
Polar angle measurements within a cortical visual field map can yield a quarter field
(either vertical meridian to the horizontal meridian) or hemi-field representation (upper
vertical meridian to the lower vertical meridian and vice versa). At the edges of one visual
field, the cortical representation of either meridian could mark one end of this map or may
reverse into the polar angle map of a neighboring field map. These reversals will be used to
locate boundaries between adjacent representations of separate cortical visual field maps
(DeYoe et al., 1994; Engel et al., 1994; Sereno et al., 1995; Engel et al., 1997a).
42
EXPERIMENT 1- Retinotopic mapping
Methods
3.1.1.2 Polar angle
The second dimension, cortical visual field maps must represent, is the eccentric
extent from fixation within the visual scene. An eccentricity map progresses from a central
visual field representation in an orderly fashion towards increasingly more peripheral field
representations. Eccentricity measurements are preferably conducted by the means of
eccentrically expanding ring stimulation at more than just two positions (Figure 3.2E).
The described temporal/spatial stimulus sequence creates a traveling wave of activity
in retinotopic cortex (DeYoe et al., 1994; Engel et al., 1994; Sereno et al., 1995; Engel et al.,
1997a; Wandell, 1999; Wandell et al., 1999). Cortical neurons respond as the spatial
configuration of the stimulus (wedge or ring) lies within their receptive field. The MR-signal
is detected as net response of these neurons in voxels of the measured brain volume at distinct
times and corresponding visual field locations. The timing of the peak response indicates the
visual field position that most effectively excites the neurons in that voxel. The response
phase for each voxel at the fundamental stimulus frequency thus encodes the preferred
stimulus angle or eccentricity within the visual field. This methodology is called phase
encoding method because the phase of the BOLD response for each voxel measures the most
effective stimulus position for that voxel and thus yields cortical representations of the visual
field. The phase encoding method has been described in detail in section 2.2.1.
3.2 Methods
3.2.1 FMRI measurements of retinotopy
In order to achieve high precision retinotopic mapping, a stimulation protocol was
applied stimulating essentially eight separate positions of the visual field in the angular as
well as the eccentricity domain. The phase information at each voxel is subsequently analyzed
(Engel et al., 1997b; Dougherty et al., 2003; Brewer et al., 2005; Liu & Wandell, 2005;
Wandell et al., 2005). The wedge and ring stimulation used in the reported retinotopic
measurements is described in Figure 3.2.
43
EXPERIMENT 1- Retinotopic mapping
Methods
Figure 3.2: Angular and eccentricity stimuli used in retinotopic mapping
Cortical polar angle representations are measured using wedge stimuli as in B. The 45° wedge is
presented at eight different positions in the visual field indicated by ciphers. A and C represent
pseudocolor legends used in subsequent figures and flat maps to indicate the preferred stimulus
location of respective cortical locations (voxels) with regard to polar angle. For instance panel A
indicates stimulus presentation in the left visual field. Cyan represents the upper vertical meridian, blue
the left horizontal meridian and magenta the lower vertical meridian. Panel E shows the maximal
extent of eight rings used to measure cortical eccentricity representation. Ciphers indicate the
respective radius of the used rings. D and F represent pseudocolor legends used in subsequent
figures and flat maps to indicate the preferred stimulus location of respective cortical locations with
regard to eccentricity. For instance panel F indicates ring stimulation in the right visual field. Red and
yellow code central stimulus preference while green and blue represent more peripheral targets.
3.2.1.1 MR-protocols:
Retinotopic measurements were conducted using the neuro-optimized GE 1.5T Signa
Horizon LX (General Electric Medical Systems, Milwaukee, Wisconsin, USA) and a 3T
whole-body Magnetom Trio (MRI systems, Siemens Erlangen, Germany).
Two anatomical brain volumes were acquired for each subject and retinotopic
measurement. Firstly a 3D high-resolution full head volume, using a fast spoiled gradient
echo sequence (FSPGR), supplying excellent white/gray matter contrast in addition to high
spatial resolution, was obtained. In the 1.5T system the high-resolution volume subtended 124
sagittal slices with voxel dimensions of 0.98x0.98mm and 1.5mm slice thickness; (TR=24s,
TE=MinFull, alpha=30°, FoV=25 cm). The high-resolution scan in the 3T system was
acquired using a 3D-magnetization prepared rapid gradient echo sequence (MP-RAGE) and
subtended 192 sagittal slices with 1mm isotropic voxels (TR=2.5s, TE=4.77ms, TI=1100ms,
alpha=7°, FoV=25.6 cm). These whole head high-resolution scans were measured only once
per subject and scanner, and served as an individual anatomical standard space for all
subsequent functional measurements.
The second additional anatomical volume with lower resolution but faster acquisition
time was obtained using a spin echo T1-weighted sequence (SET1), measuring 20 slices,
oriented orthogonal to the calcarine sulcus and identical slice orientation as the functional
volumes in the 1.5T system, while in the 3T system a 3D-fast low angled shot sequence
(FLASH) was used, yielding 0.7 x 0.7mm voxel dimensions with 3mm slice thicknes
(TR=1.5s, TE=4.9ms; alpha= 25°; FoV=180x180mm).
44
EXPERIMENT 1- Retinotopic mapping
Methods
Functional measurements in the 1.5T system were obtained using an EPI-gradient
recalled sequence, yielding 144 functional brain volumes with 20 slices of 2.2mm slice
thickness and 3x3mm voxel dimensions (TR=2s, TE=35ms, alpha=80°, FoV=130mm).
Functional measurements in the 3T system were obtained using also BOLD sensitive EPIsequence received in a 5“ surface coil (RAPID Biomedical GmbH, Würzburg, Germany)
placed at the occipital pole yielding 20 slices with 2.81 x 2.81mm voxel dimensions and 3mm
slice thickness (TR=2s; TE=30ms; alpha=80°, FoV=180x180mm) covering the occipitoparietal cortex. The initial four scans were discarded.
Polar angle and eccentricity measurements per subject were repeated three times. The
respective black and white checkerboard pattern was presented for 24 seconds containing 8
phase steps. This cycle was presented 12 times per run. The stimuli were projected onto a
screen in the scanner room using a Sharp XG-SV1E projector with a resolution of 800x600
pixels, mean luminance of 1175 cd/m² and Michelson-contrast ((max-min)/(max+min)) of
98%. Subjects viewed the stimuli by means of a mirror system within the scanner while
fixating a central marker. The largest eccentricity ring subtended 14° visual angle while each
single ring step obtained 1.75° of the visual field. Each of the polar angle wedges measured
45°.
Time series for every voxel underwent the phase encoding method described in detail
in section (2.2.1.) which yields amplitude and phase of the periodic BOLD-response for all
frequencies. Only voxels whose time series correlated no less than 0.3 with the fundamental
stimulus frequency were included into retinotopical analysis of phase reversals and
eccentricity measurements. Flat maps in Figures 3.6 to 3.9 depicting polar angle and
eccentricity measurements are thresholded at that value. Flat maps representing MT-localizer
stimulus activity are presented at 0.5 coherence threshold. The MT-localization procedure is
explained in the next section.
3.2.2 MT+ localization
As investigations regarding color and motion dependence on the responses of visual
cortex are conducted we chose to include a region known to be highly sensitive to moving
stimuli. For that reason we functionally map a cortical region, exhibiting high motion
sensitivity. The area is known as MT-complex (MT+), named after presumed homologous
motion sensitive regions in the medial temporal (MT) and medial superior temporal (MST)
cortex in the macaque monkey. In humans, this region lies on the lateral surface of the
occipital lobe, mainly within the dorsal/posterior limb of the inferior temporal sulcus (ITS)
(Zeki et al., 1991; Watson et al., 1993; Tootell et al., 1995a; Dumoulin et al., 2000). Relative
45
EXPERIMENT 1- Retinotopic mapping
Methods
to other visual areas, MT+ shows increased sensitivity to low-contrast moving stimuli (Tootell
et al., 1995a) and its activity underlies direction-selective neurons (Heeger, 1999; Huk &
Heeger, 2000; Huk et al., 2002). Chromatic signals in MT+ have been reported (Seidemann et
al., 1999; Thiele et al., 1999; Wandell et al., 1999; Liu & Wandell, 2005) but while the
relative proportion compared to achromatic or luminance contribution is controversial there is
strong evidence that MT+ plays an important role in motion perception (Salzman et al., 1990;
Newsome et al., 1991; Shadlen et al., 1996).
3.2.2.1 FMRI measurements of MT+
Localization of MT+ in human visual cortex in this thesis, was based on responses to
stimuli that alternated in time between moving expanding/contracting- and stationary dot
fields, the conventional method (Zeki et al., 1991; Watson et al., 1993; Tootell et al., 1995b;
Huk et al., 2002) in addition to identification of angular and eccentricity representation within
that region. The dot fields subtended 15° visual angle, had approximately 10% Michelsoncontrast and were presented on a mean gray background. The dot fields moved towards and
away from fixation for 12 seconds alternating their direction every 3 seconds followed by 12
seconds of stationary dot field presentation. Subjects were to fixate a central marker while
attending to the dot fields. To keep the attentional load constant, subjects were asked to
indicate a sudden change in speed of the dots by means of a button press as stimuli moved
while in the stationary condition a sudden contrast change of the dots was to be indicated.
This procedure was repeated 20 times per run with a TR of 2s yielding 240 brain volumes per
run. Two functional runs were measured per subject. Anatomical scans regarding the two MRsystems were identical to the retinotopic measurements protocol for both MR-systems (see
section 3.2.1.1), while the high-resolution anatomical scan was omitted. Voxelwise time series
were analyzed using the phase encoding procedures at fundamental stimulus frequency of
1/20 cycles per scan, as described above (section 2.2.1.). Correlation values were thresholded
at 0.5 for measurements of both MR-systems and projected onto flattened maps of the
segmented gray matter sheet of the occipital cortex. Flat map generation is described in the
next section.
3.2.3 Cortical flat maps
In order to visualize average data from multiple slices, a flattened representation (flat
map) of the occipital gray matter was produced per subject and hemisphere. On flattened 2D
surfaces, the spatial relationship between cortical visual areas can be inspected much more
precisely as compared to multiple slices of 3D brain volumes, because medial, ventral and
46
EXPERIMENT 1- Retinotopic mapping
lateral
views
are
simultaneously
Results
visible.
The
mrVista
Toolbox
was
used
(www.white.stanford.edu/software/). The flatmaps were generated passing through essentially
four steps, (Teo et al., 1997), identical for measurements in both MR-systems.
Initially, a high-resolution anatomical T1-weighted 3D whole head scan (FSPGR) was
segmented manually once for each hemisphere respectively using mrGray (mrVista). The
high-resolution anatomical brain volume served as anatomical reference for all subsequent
functional measurements. The segmented gray matter sheet was computationally flattened
using mrFlatMesh (mrVista). An additional anatomical volume (SET1) was acquired for
every functional session and subject with the same slice position as the subsequent functional
scans. The anatomical SET1 brain volume was rotated onto the high resolution FSPGRanatomy, yielding rotation- and translation parameters. The rotation was performed using the
software package mrAlign2, which is part of the mrVista toolbox. Since SET1 and functional
EPI volumes were acquired using the same slice orientation, identical rotational and
translational parameters could be applied to the functional volumes. It is thus possible to
transform the coordinates of activated voxels for a given visual stimulation from functional
brain volumes into high resolution anatomical coordinates which can then be reliably
transferred to flattened cortical maps. By doing so, it is assured, that representation of cortical
activity on flat maps originates from gray matter. The final step contains a combination of
semiautomatic software (mrFindBorders) and manually inspecting the phase reversals on
cortical representations of borders of the visual areas, generated by polar angle stimulation,
and simultaneously taking activity patterns, generated by eccentricity stimulation, into
account.
Flat maps representing activity elicited by polar angle and eccentricity stimuli as well
as MT+ localization in two MR-systems will be compared in the next section for two subjects.
Furthermore, two additional subjects will be presented regarding their cortical organization of
visual areas, measured in the 3T system.
3.3 Results
3.3.1 Polar angle measurements
V1 represents a homogenous visual contralateral hemifield within the entire calcarine
sulcus extending from a lower vertical meridian representation to an upper vertical meridian
representation found on the upper and lower lip of the calcarine sulcus respectively as
illustrated in Figure 3.3F (Dougherty et al., 2003; Wandell, 2004; Brewer et al., 2005; Tyler
et al., 2005). These two V1 meridian representations are shared on the ventral side with V2
47
EXPERIMENT 1- Retinotopic mapping
Results
ventral (V2v) and on the dorsal side with V2 dorsal (V2d). Each of these V2 maps represents
a visual quarterfield, which, taken together, represent an entire hemifield. The quarter field
representation for both regions ends at a horizontal meridian representation, which is shared
by quarterfield maps of V3 ventral (V3v) and V3 dorsal (V3d) on the ventral and dorsal side
respectively. Both maps, V2 and V3 are therefore regarded as discontinuous maps. Logically,
the anterior dorsal border of V3d represents a lower vertical meridian, while the anterior
ventral border of V3v represents an upper vertical meridian (Dougherty et al., 2003). These
spatial relationships were found in all of the hemispheres tested and can be scrutinized in the
left columns of flatmap graphs in Figures 3.6 to 3.9 and in Figure 3.3. Every hemisphere
investigated displayed a full hemifield representation, sharing an upper vertical meridian with
V3v and extending through a horizontal meridian representation towards a lower vertical
meridian representation as measured using polar angle stimuli (Figure 3.2). This full hemifield
representation extends from the posterior end (upper vertical meridian) towards more anterior
portions (lower vertical meridian) of the fusiform gyrus and occupies lateral parts of the
collateral sulcus in some hemispheres (compare with Figure 3.1).
The first full ventral hemifield representation beyond V3v was highly regular across
subjects and resulted in hV4 mapping. The first full dorsal hemifield representation beyond
dorsal V3 draws a somewhat different picture. The anterior border of dorsal V3 constitutes a
lower vertical meridian representation. In order to map a full hemifield cortical representation
should anteriorly extend through a horizontal meridian followed by a representation of an
upper vertical meridian.
Anterior to V3 dorsal we find in fact a clear upper vertical meridian following a
horizontal meridian representation. This upper vertical meridian representation appears,
however discontinued in most of the hemispheres or is missing a posterior part in some
hemispheres. Flat maps of the left and right hemisphere of subject Jm05 in Figure 3.6
illustrate this discontinuity. The upper vertical meridian representation follows a relatively
parallel path to the V3 dorsal border starting somewhat near to what can be expected to be the
confluent V1-V3 central (foveal) representation. However, following that path towards more
peripheral representations (compare to eccentricity scans) this upper vertical meridian
representation is interrupted somewhere midway. The gap coincides with at least one foveal
representation on the dorsal lateral surface as the eccentricity map is inspected (middle
columns). The same pattern is observed for both hemispheres in that subject.
In some hemispheres, at coherence threshold of 0.3, our polar angle measurements
revealed no upper vertical meridian representation oriented roughly parallel to the posterior
V3 border, but instead another lower vertical meridian was found, essentially representing a
48
EXPERIMENT 1- Retinotopic mapping
Results
quarter field (see for instance left hemisphere, subject Dn45 in Figure 3.7). Some hemispheres
on the other hand exhibit obvious representation of an upper meridian at both posterior and
anterior locations (see for instance left polar angle maps for subject Jm05 in Figure 3.6).
A
B
D
Left
E
F
ei00
C
V3A
MT+
G
V3d
V2d
V1
hV4
V2
V3
Figure 3.3: Polar angle maps identify borders of visual areas
A and B show a lateral and medial 3D rendering of subject ei00s full left hemisphere respectively.
Panel D and F magnify the lateral and medial view of the subjects occipital cortex respectively. Panel
A, B, D and F: Polar angle representation are overlayed and restricted to the region of interest
coordinates defined in E. Panel E shows clear retinotopic organization of the left occipital cortex of
ei00 generated by means of polar angle stimulation in the right visual field (see pseudo-color-map, left
lower corner in E). Magenta represents upper vertical meridian, blue represents horizontal meridian
and cyan lower vertical meridian. Solid colored lines in panel E mark the borders of the visual areas as
identified by phase reversals. Color codes for visual areas remain identical throughout this thesis.
V1~>blue; V2d/V2v ~>red; V3d/V3v ~>yellow, hV4 ~> green, V3A ~> magenta and MT+ ~>cyan. ROIs
are overlayed onto the lateral and medial surface of ei00’s occipital cortex respectively in panel C and
G. Besides V1 three additional continuous hemifield representations could be precisely mapped in
several regions of occipital cortex. A clear hemifield map on the ventral surface, extending from an
upper vertical meridian representation at the border to V3v to a lower vertical meridian representation
was classified as hV4. On the dorsal surface a clear hemifield map extends from a lower vertical
meridian representation at the border to V3d to an upper vertical meridian representation anteriorily
and was classified as V3A. In addition, a spatially distinct but quite obvious hemifield representation
was identified on the lateral surface. MT+ localization scans (see Figure 3.5) confirmed this region as
human motion selective area MT+. These additional maps could be conformed in several other
subjects (see Figures 3.6 to 3.9).
3.3.2 Eccentricity measurements
Eccentricity measurements revealed clear continuous representation for several visual
areas (V1, V2 V3, hV4). Starting with a large central (foveal) representation at the posterior
end of the calcarine sulcus extending towards increasingly anterior positions along the medial
surface representing increasingly more peripheral positions in the visual field. Eccentricity
representations form a semicircular pattern on flat maps, which can be fully appreciated along
the middle columns of the flat map graphs in Figures 3.6 to 3.9 and in Figure 3.4. Eccentricity
representation in hV4 is however clearly distinguishable from V1-V3 with respect to more
peripheral stimulus locations in the visual field. While V1 to V3 represent the entire
49
EXPERIMENT 1- Retinotopic mapping
Results
contralateral visual hemifield, hV4 exibits a bias for more central targets. This can be
observed on all eccentricity flat maps in the middle columns of Figures 3.6 to 3.9 and in
Figure 3.4. While V1-V3 represent peripheral targets up to the largest eccentricity measured
(14° visual angle) coded by the blue pseudocolor, the hV4 map does not exceed the green
pseudocolor, representing approximately 10° of visual angle.
Figure 3.4: Eccentricity representation
A
B
Panel A: ventral medial view onto the full left
hemisphere of subject ei00. Panel B: Magnification of
the ventral medial view. Panel C: left occipital flat
map for subject ei00. The color codes for regions of
interest are the same as in Figure 3.3. The
pseudocolor map in the left lower corner of panel C
indicates the position of the visual field. The maximal
eccentricity was measured at 14° visual angle.
left
C
ei00
Recent reports suggest to separate what appears to be a full hemifield beyond V3 dorsal,
preferably at a dorsal foveal representation, displaced from the confluent fovea (Smith et al.,
2001; Tyler et al., 2005; Wandell et al., 2005). This procedure will assign the smaller, more
posterior part of this hemifield representation to V3B, leaving the larger, anterior part to V3A
while both still represent a full hemifield. A clear central (foveal) representations on the dorsal
surface could not be mapped in all subjects and is for instance completely missing in the left
hemiphere of subject Ei00. Others propose one homogenous hemifield map in that region like
its putative macaque counterpart (Tootell et al., 1997).
In order to use equal criteria regarding the border of the dorsal visual hemifield map
for all subjects, we chose to guide our definition of the anterior border of the dorsal hemifields
along the upper vertical meridian representation wherever present to derive a complete
hemifields representation within that area. For simplicity we adapted a naming convention set
50
EXPERIMENT 1- Retinotopic mapping
Results
forth by Tootell (Tootell et al., 1997) describing a human dorsal hemifield adjacent to dorsal
V3 as V3A, homlogous to macaque. Our definition yields a dorsal representation of the entire
contralateral hemifield along the borders of dorsal V3 in all subjects. The largest fraction of
this region will contain V3A but also V3B, satisfying newer conventions (Brewer et al., 2005;
Tyler et al., 2005; Wandell et al., 2005).
3.3.3 MT+ retinotopy and localizer measurements
Experimental procedures aiming at motion selective cortex in addition to retinotopic
mapping permit to precisely localize MT+. Polar angle measurements on the lateral occipital
surface revealed clear retinotopical organization. We confirm a full contralateral hemifield
representation mostly in the ascending branch of the posterior portion of the inferior temporal
sulcus. Upper and lower vertical meridian representation were robustly measured in every
hemisphere exceeding a coherence threshold of 0.3.
Figure 3.5: MT+ Localization
Panel A: Lateral view of full left hemisphere
of subject ei00. Panel B: Magnification of
lateral occipital cortex of subject ei00. Panel C:
Flat map of left hemisphere. The location
coordinates and color codes for regions of interest
are the same as in Figure 3.3. All panels show cortical
regions where voxel time series correlated stronger
than 0.5 coherence threshold to moving versus static
dot field stimulation (see text for details). Note multiple
regions are highly activated by motion stimuli.
MT+ localization based solely on moving versus static
targets is not sufficient. Polar angle and eccentricity
information must be taken into account. Experiment 2
will suggest an alternative approach.
A
B
C
V3A
1
0.5
0
r
MT
hV4
V3d
V2d
V1
V3v V2v
Eccentricity measurements revealed a more variable organization of MT+. Applying a
high coherence threshold of 0.3 decreased activity representation in some hemispheres, while
wedge stimulation gave still robust signals. However compared to the full ventral hemifield
representation (hV4), the lateral hemifield representation (MT+) responds to largest
eccentrical stimuli while hV4 does not. This can be best appreciated in Figure 3.6 and 3.8 for
subjects Jm05 and Dn45.
51
EXPERIMENT 1- Retinotopic mapping
Results
A
3T
left
V3A
V3d
V2d
MT
V1
r
1
hV4
0.5
V3v V2v
0
right
V3A
V3d
V2d
1
0.5
MT
hV4
r
V1
V2v V3v
0
Figure 3.6: Retinotopic organization and MT+ localization for subject Jm05 (3T)
Left column: phase reversal maps for rotating wedge stimulation (Coherence Threshold 0.3). Middle
column: eccentricity maps for expanding ring stimulation (Coherence threshold 0.3). Right column:
MT+ localization measurements expanding/contracting dot fields vs. stationary dotfields (Coherence
threshold > 0.5, see legend). Insets in lower left corner of left and middle column panels indicate
stimulus position in the visual field. Colored solid lines mark the borders of visual areas.
right
V3A
V3d
V2d
1
0.5
r
MT
V1
V2v
V3v hV4
0
Figure 3.7: Retinotopic organization and MT+localization in right hemisphere of
subject Ei00 (3T).
Left hemisphere flat maps can be seen in Figure 3.3, 3.4, and 3.5. Plotting conventions as in Figure
3.6.
52
EXPERIMENT 1- Retinotopic mapping
1.5 T
Results
left
V3A
V3d
V2d
MT
V1
r
1
V2v
0.5
hV4
0
right
V3v
V3A
V3d
V2d
V
1
r
1
V2v
0.5
3T
0
MT
V3v hV4
left
V3A
V3d
V2d
MT
1
V1
r
V2v
0.5
hV4 V3v
0
right
V3A
V3d
V2d
V1
1
0.5
0
MT
r
V2v
V3v
hV4
Figure 3.8: Retinotopic organization and MT+localization for subjects Dn45 in
1.5T and 3T-MR-systems.
First and third row left hemisphere. Second and forth row right hemisphere. First and second row 1.5T,
third and fourth row 3T-measurements. All other plotting conventions as in Figure 3.6.
53
EXPERIMENT 1- Retinotopic mapping
1.5 T
Results
left
V3A
V3d
V2d
1
V1
MT
r
V2v
hV4
0.5
V3v
0
right
V3d
V2d
V3A
MT
V1
V2v
r
1
hV4
V3v
0.5
0
3T
V3A
left
V3d
V2d
V1
r
1
V2v
V3v
MT
hV4
0.5
0
right
V2d V3d
1
0.5
r
V3A
V1
V2v
V3v
MT
hV4
0
Figure 3.9: Retinotopic organization and MT+localization for subjects Lk13 in
1.5T and 3T-MR-systems.
First and third row left hemisphere. Second and forth row right hemisphere. First and second row 1.5T,
third and fourth row 3T-measurements. All other plotting conventions as in Figure 3.6.
54
EXPERIMENT 1- Retinotopic mapping
Discussion
The orientation of the foveal/peripheral visual field representation of MT+ with
respect to the other visual areas on the unfolded cortical surface appears quite variable
between subjects, but data from different magnets for the same subjects reveal the same
organization principle. We attribute this observation to variable degrees of folding of the gray
matter sheet for each subject, which will result in variable positions on the flat map, as well as
probable different hemodynamic delays (Tyler et al., 2005).
3.4 Discussion
Human medial occipital representations of the visual field have been under substantial
investigation, especially in the field of neuroimaging. Using fMRI, three robust visual
hemifield maps, namely V1, V2 and V3 are routinely localized and well established (DeYoe
et al., 1994; Engel et al., 1994; Sereno et al., 1995; Engel et al., 1997a; Dougherty et al.,
2003; Wandell et al., 2005). In addition to these three maps we report precise localization of
the borders of visual areas hV4, V3A and MT+, which as V1, contain continuous
representation of the full contralateral visual hemifield.
While the location and spatial extent of primary visual areas (V1-V3) are well
documented (Dougherty et al., 2003; Brewer et al., 2005; Wandell et al., 2005) cortical
representations of full hemifields beyond primary cortex are subject to intense controversy.
Probably most contentiously discussed in the recent literature is the spatial representation of
the visual field in the first ventral surface map, V4.
Guided by macaque anatomy, Tootell and Hadjikhani set out to clarify whether human
V4 is split into ventral and dorsal components which would entail, each of those regions to
represent only a respective quarterfield of the contralateral visual field (Tootell, 2001). It is,
however, pointed out by this group, that the dorsal region, which they suspected to represent
dorsal V4, lacks many features that would qualify it as the dorsal quadrant of V4 ventral,
making a ventral or dorsal quarterfield representations in V4 unlikely (Tootell, 2001; Tyler et
al., 2005). A recent report persuasively showed that the first ventral map beyond V3v is
indeed a full hemifield representation, bordered anteriorly by a lower vertical meridian
representation (Brewer et al., 2005). Our data is in concordance with these proposals as all
hemispheres measured in this work show a full hemifield representation with a lower vertical
meridian running roughly parallel to the V3 ventral (Wandell et al., 2005).
In addition it is generally agreed that hV4 shares the confluent fovea with V1-V3
while dorsal area V3A/B may have its own foveal representation, displaced to the large foveal
confluence (Press et al., 2001; Smith et al., 2001; Brewer et al., 2005; Wandell et al., 2005).
The measured eccentricity maps support the foveal hV4 representation to be confluent with
55
EXPERIMENT 1- Retinotopic mapping
Discussion
the V1-V3 foveal representation in all hemispheres shown. On the dorsal cortical surface
however the majority of hemispheres reveal a foveal representation distinct from the confluent
foveal representation (see Figures 3.6 to 3.9 and 3.4) (Tyler et al., 2005). These foveal
representations are thought to be shared by two full hemifield representations beyond V3d,
namely V3A and V3B (Tyler et al., 2005; Wandell et al., 2005). Since this non-confluent
dorsal foveal representation was not observed in all hemispheres these putative separate
hemifield representations have been joined to form the, what we named V3A hemifield map,
in this work. Future imaging research, focusing on the dorsal surface of human cortex is
necessary to clarify the existence and spatial relationship of dorsal foveal representations.
Scrutinizing the eccentricity maps again reveals besides the confluent V1-V3 foveal
representation, two additional ventral foveal representations in all hemispheres and two
additional dorsal foveal representations in some hemispheres well separated from the
confluent fovea (best seen in Figure 3.6: left hemisphere for subject Jm05; Figure 3.7: right
hemisphere for subject Ei00 and Figure 3.8: left hemisphere subject Dn45). The posterior but
displaced ventral foveal representation has already been associated with ventral occipital area
VO-1, being situated anterior to hV4 (Brewer et al., 2005). A simiar organization may be
found on the dorsal surface, in which the first displaced foveal representation, seen in most of
the eccentricity maps in our data, could be associated with the hemifield representation of
V3A, while V3B may share the confluent foveal representation of V1-V3 and hV4.
Finally, an additional ventral occipital full hemifield representation (VO-2) an
associated foveal representation was found anterior to VO-1 (Brewer et al., 2005). A similar
organization on the dorsal surface may correspond to a full hemifield representation known as
V7 (not shown in our flat maps), having its own foveal representation (Tootell et al., 1998;
Press et al., 2001; Tyler et al., 2005). A similar symmetrical canonical scheme has been
suggested by Tyler et al. (Tyler et al., 2005).
However, the percentage of eccentrical representation varies substantially within and
across visual areas. The cortical magnification factor leads to a cortical representation of the
fovea of about 8% in V1 while the area of highest acuity obtains only 0.01% of the entire
retina (Sharpe & Nordby, 1990). This may partly be explained by the increased density of the
retinal photoreceptors in that region as compared to the peripheral density (Wässle et al.,
1990), while the central visual field map in primate V1 may be magnified beyond what may
be derived from foveal receptor density (Inouye, 1909; Talbot & Marshall, 1941; Daniel &
Whitteridge, 1961; Perry & Cowey, 1985; Horton & Hoyt, 1991a; Dougherty et al., 2003).
The motion responsive region MT+ on the lateral occipital surface has been
characterized as having coarse retinotopically organized maps, while the subdivision MST of
56
EXPERIMENT 1- Retinotopic mapping
Discussion
the MT+ complex may not be retinotopically organized (Huk et al., 2002). Polar angle
representations in the inferior temporal sulcus reported here replicate previous finding and
securely map MT+ within that region. Every hemisphere showed clear upper and lower
vertical meridian representation. The upper vertical meridian representation was located more
anteriorly than the lower vertical meridian representation in every hemisphere, confirming a
prior report (Huk et al., 2002). Eccentricity organization in area MT+ have been characterized
to represent more peripheral targets in the superior part of MT+ while central targets are being
represented in the inferior part (Huk et al., 2002). We find the same organization in most
hemispheres, even though the left hemisphere of subject Jm05 exhibits reverse eccentricity
representation (see Figure 3.6). In general, the local cortical anatomy of the MT+ region is
quite variable across individuals (Watson et al., 1993; Tootell et al., 1995b; Dumoulin et al.,
2000).
This chapter summarized experimental procedures leading to definition and
determination of visual field map representations. We securely identified visual field maps in
V1, V2 V3, hV4, V3A and MT+. These regions will serve as an outline to define regions of
interest in subsequent experiments of this thesis. The next experiment will investigate speed
dependence on the BOLD-response in these visual field maps using equal cone-contrast
stimuli.
57
EXPERIMENT 2- Speedtuning
Introduction
4 EXPERIMENT 2- Speedtuning
4.1 Introduction
How neural activity relates to the experience of moving colors is controversial,
because key behavioral results remain at odds with existing neurophysiological data.
Experiment 2 aimed to provide speed-tuning properties for several visual areas that suggest an
association with psychophysically derived motion perception mechanisms. In addition the
effects of speed at a suprathreshold contrast level for five color directions (L+M, L-M, S(L+M), L+M+S, S) will be investigated within these areas. Best separating stimuli between
area hV4, representing the ventral path, and area MT+, representing the dorsal path, will be
provided, to facilitate future research regarding strict segregation of these paths. The results
will be discussed in the context of contemporary neurophysiological and imaging studies,
addressing color and speed tuning in human- and monkey visual cortex as well as the long
lasting controversy regarding the magnitude of color-opponency in human visual cortex.
4.1.1 Luminance - detection vs. discrimination
For moving luminance gratings, contrast thresholds for detection and discrimination of
direction of motion are highly similar at all temporal frequencies and for all eccentricities
within 5 deg of the fovea (Levinson & Sekular, 1975; Watson et al., 1980; Lindsey & Teller,
1990; Derrington & Henning, 1993; Metha et al., 1994; Gegenfurtner & Hawken, 1995;
Stromeyer et al., 1995). Interpretations of these results yield the idea, that the same neural
mechanism is employed for detection and discrimination of luminance motion and that this
mechanism is direction-selective (Watson & Robson, 1981).
4.1.2 Isoluminance - detection vs. discrimination
On the other hand, psychophysical experiments investigating isoluminant motion
perception draw a different picture. These reports have shown that, although motion
perception is possible at isoluminance (Cavanagh et al., 1984; Lindsey & Teller, 1990;
Cavanagh & Anstis, 1991), the perceived speed is slower than for cone-contrast matched
luminance stimuli (Cavanagh et al., 1984; Mullen & Boulton, 1992).
For stimuli with low temporal frequencies, presented foveally, the motion detection
sensitivity is better than the sensitivity for motion discrimination. This difference increases at
more peripheral locations and is more pronounced at higher temporal frequencies
(Gegenfurtner et al., 1994; Gegenfurtner & Hawken, 1995, 1996).
58
EXPERIMENT 2- Speedtuning
Introduction
4.1.3 Two motion mechanisms
The discrepancy between detection and discrimination of isoluminant motion has led
to the assumption of two motion processing streams that differ mostly in their temporal
characteristics and are essentially sensitive to color (Hawken et al., 1994). Gegenfurtner and
Hawken (1996) distinguish a fast channel from a slow channel. The fast channel veridically
represents the velocity of moving patterns (also at isoluminance) but might not code the color
of such patterns. Conversely, the slow channel is highly sensitive to color and signals the
direction of slowly moving patterns while the coding of stimulus velocity is not veridical
(Gegenfurtner & Hawken, 1996). The neuronal substrates of these two different functional
channels in humans are unknown.
The main question being addressed in the second experiment is consequently, which
cortical area(s) can be attributed to these two motion channels derived psychophysically. In
the course of action we may additionally ask, if chromatic response differences within and
between visual areas, found in previous fMRI studies and described in section 1.2.3.6, change
depending on speed (Kleinschmidt et al., 1996; Engel et al., 1997b; Zeki & Marini, 1998;
Wandell et al., 1999). A third question will address chromatic response differences as we
investigate hierarchically higher tier visual areas?
Gegenfurtner and Hawken suggest area MT as a candidate for the fast channel because
it is of paramount importance in the processing of visual motion (Zeki, 1974; Movshon et al.,
1985), smooth-pursuit eye movement (Newsome et al., 1985), is highly direction selective
(Newsome et al., 1991) and responds to isoluminant stimuli (Saito et al., 1986; Charles &
Logothetis, 1989; Dobkins & Albright, 1994; Gegenfurtner et al., 1994; Wandell et al., 1999).
MT can however not be held responsible for the slow channel, because contrast sensitivity in
MT cells (for chromatic stimuli) is not sufficient for the excellent behavioral sensitivity to
slowly moving isoluminant gratings (Gegenfurtner et al., 1994) (see experiment 3).
Further evidence supporting separate speed dependent motion channels comes from a
patient suffering from bilateral damage to the lateral temporo-occipital cortex, resulting in
cerebral akinetopsia. Motion perception in this patient was only impaired for fast moving
stimuli while no deficit could be reported for slowly moving stimuli below 4Hz (Zihl et al.,
1983; Hess et al., 1989). Residual responses to motion stimuli in area V3 could be observed in
this patient (Shipp et al., 1994).
Gegenfurtner and Hawken (1996) suggest a network comprised of V3 and V4 for the
slow channel, since single-cell recordings in macaque V3 revealed joint selectivity for color
and motion (Gegenfurtner et al., 1997), while V4 is well known for its color sensitivity (Zeki,
1983b; Kleinschmidt et al., 1996; McKeefry & Zeki, 1997; Zeki & Marini, 1998; Beauchamp
59
EXPERIMENT 2- Speedtuning
Methods
et al., 1999; Wade et al., 2002; Brewer et al., 2005) and is sensitive to the direction of motion
(Desimone & Schein, 1987; Tolias et al., 2001a; Tolias et al., 2005b). In addition, a number
of studies have shown that chromatically opponent mechanisms most likely underlie both the
detection and discrimination of moving chromatic targets (Krauskopf & Farell, 1990; Metha
et al., 1994; Gegenfurtner & Hawken, 1995; Stromeyer et al., 1995).
To investigate an association of cortical regions to the differential properties of the
proposed fast and slow motion perception channels, two assessment criteria should be tested.
The first criterion pertains to temporal frequency sensitivity and will be tested in the present
experiment. The second criterion pertains to differential chromatic contrast sensitivity
between the two motion channels and will be part of the measurements conducted in
experiment 3.
As for the first criterion, temporal frequency sensitivity differences, a visual area
corresponding to the highly-contrast-sensitive, color-opponent slow mechanism should prefer
chromatic over luminance stimulation at slow speeds (< 4Hz) and responses should decrease
as temporal frequencies exceed 4Hz (Gegenfurtner & Hawken, 1996). By contrast, an area
belonging to the fast motion pathway should consequently be equally responsive to all
stimulus velocities regardless of color direction and responses should not decline at
frequencies above 4Hz (Gegenfurtner & Hawken, 1996).
Experiment 2 will present speed tuning properties for retinotopically mapped visual
areas V1, hV4, V3A and MT+ elicited by stimuli presented at equal cone-contrast of 14%
along five directions of cone-contrast space and three velocities. These response patterns
suggest MT+ as neural substrate for the fast channel while hV4- and V3A- response patterns
allow association with the slow channel.
4.2 Methods
4.2.1 Subjects
Six human subjects participated in fMRI measurements. Four were naive to the
purpose of the study. All subjects had normal or corrected to normal vision and gave their
written informed consent prior to the experiment. The ethics committee of the University of
Magdeburg approved the experiment.
60
EXPERIMENT 2- Speedtuning
Methods
4.2.2 Visual display
For the fMRI-measurements the stimulus presentation was controlled using Cogent
2000 (available at www.vislab.ucl.ac.uk) under Matlab 6.5. (Math Works, Natick, MA) on a
pentium 4 Desktop Computer with an integrated WinFastA350 Graphics Card and displayed
on a flat-panel TFT-display (NEC, MultiSync LCD 2180UX, Ithaca, Illinois) with a
resolution of 1600x1200 pixels and a screen refresh rate of 60 Hz. The TFT display was
inside a front-transparent shielded box in the scanner room. Subjects viewed the display
through binoculars.
The spectral properties of the displays were characterized with a pr650 Photoresearch
(Chatsworth, CA) spectroradiometer. The cone-contrast stimuli were calculated with the 10°
Stockman and Sharpe sensitivity functions (Stockman & Sharpe, 1998, 2000).
The temporal stability of the display was assessed for all our stimuli by measuring the
photoelectric current elicited in a photodiode. We found no signal attenuation for any contrast
direction across the different speeds.
4.2.3 FMRI protocol
Functional magnetic resonance imaging was performed on a 3T whole-body MRI
system (Siemens Magnetom Trio, Erlangen, Germany) with a BOLD-sensitive EPI-sequence
(TR=1500ms; TE=30ms; alpha=80°) using a 5“ surface coil (RAPID Biomedical GmbH,
Würzburg) placed at the occipital pole. Twenty slices (matrix size 64 x 64 voxels; spatial
resolution: 2.81 x 2.81 x 3mm; no gap) were oriented approximately perpendicular to
calcarine sulcus and covered the occipito-parietal cortex. An additional T1-weighted
anatomical volume was acquired (3D-FLASH-Sequence: matrix: 256 x 256; FoV: 180 x 180
mm; spatial resolution: 0.7 x 0.7 x 3mm; TE=4.9 ms; TR=15000ms, alpha: 25°) with the same
slice positions as the functional images. This anatomical image served to align the functional
data with a whole head high-resolution anatomy volume (3D-MP-RAGE-Sequence: matrix:
256 x 256; 192 slices per volume; FoV: 256 x 256mm; spatial resolution: 1 x 1 x 1mm;
TR=2500 ms; TE=4.77ms; TI=1100ms; alpha: 7°). This high-resolution scan served as an
individual standard space in which the results from the functional scans could be referenced to
the location of visual areas obtained by retinotopic mapping as described in section 3.2.1.
4.2.4 Stimulus & experimental design
A rotating radial sinusoidal grating was presented on a mean gray background in an
annulus (inner radius ~0.5°, outer radius ~10° visual angle) centered on a fixation marker
(Figure 2.1). The grating contrast was modulated along 5 directions in cone-contrast-space
61
EXPERIMENT 2- Speedtuning
Methods
(L+M, L-M, S-(L+M), L+M+S and S, with (L= ΔL/L, M=ΔM/M S=ΔS/S) and rotated at 3
different temporal frequencies (1, 4 and 10Hz, cycles per second). The direction of motion
changed periodically. Cone-contrast (i.e. the length of the stimulus vectors in cone-contrastspace, RMS) was held constant in all color directions at 14%, the maximum contrast in the (LM) direction in which our display produced the least cone-contrast out of all five directions.
The stimuli were presented in a balanced block design (each condition having the same priorblock-history). The order of conditions was changed between runs. All subjects performed 7
runs of 16 minutes.
To control for attention in the fMRI experiment, subjects performed an attention
binding number discrimination task at the fixation marker throughout the runs. The runs
started and ended with 30s blocks of the attention task on a mean gray background without
additional stimulation. These blocks were used to define the baseline BOLD-level.
S
L+M+S
30s
S-(L+M)
30s
L-M
30s
L+M
bg
30s
30s
30s
Figure 4.1: A sample stimulus presentation sequence.
Each run started and ended with a mean gray background for 30 seconds. A fixation task was present
throughout the run. Stimuli consisted of sinusoidal gratings, modulated along 5 directions in conecontrast-space (L+M, L-M, S-(L+M), L+M+S and S) and rotated at 3 different speeds (1, 4, and 10 Hz,
cycles/s). The rotation direction changed periodically. The stimuli were presented in a balanced block
design (each condition having the same prior-block-history) and the cone-contrast was kept constant
at 14%. Note that physical equality pertaining to cone-excitation leads to substantially different
perceptual saliency and saturation of the colors (Mullen et al., 2005).
62
EXPERIMENT 2- Speedtuning
Results
4.2.5 Analysis
Functional MR-Images were realigned and smoothed (6mm kernel, FWHM) using
spm99 (Wellcome Department of Cognitive Neurology, University College, London UK).
BOLD-signal modulation was determined for all conditions by fitting a box-car-function,
reflecting the stimulus time course, convolved with a standard hemodynamic response
function to the time-series of each voxel (see General Methods Section 2.2.2.). The result is a
weight factor (beta) that reflects the strength of BOLD modulation in each voxel for each
condition. Individual t-images were computed for each subject, revealing activation clusters in
which the cone-contrast stimuli led to significant activation compared to the baseline
condition (no annulus presented).
Areas V1, V2, V3, V3A and V4 were determined using the mrVista-toolbox (Teo et al.,
1997) and standard retinotopic mapping methods described in section 3.1.1 (Engel et al.,
1994; Sereno et al., 1995; Wade et al., 2002). Area MT+ was functionally identified based on
responses to stimuli that alternated between moving and stationary dot patterns described in
section 3.2.2 (Zeki et al., 1991; Tootell et al., 1995b; Huk et al., 2002) in addition to
retinotopic mapping (see section 3.1.1).
The visual-area-ROIs from the independent localizer scans were overlayed onto the timage to confirm the localization of the activation within the visual areas under investigation.
For the subsequent ROI based analysis the mean of responses at 27 voxel coordinates,
forming a cube, centered on the local maximum of the individual t-map-cluster that could be
safely assigned to lie within the borders of the visual areas (V1,V3A, V4, MT) was used.
These values were individually normalized for each ROI by the standard deviation over all
stimuli presented. The measure obtained that way reflects the individual BOLD-gain changes
over stimulus conditions and renders the BOLD-signal changes better comparable between
cortical sites (Logothetis, 2003). It further reduces effects of absolute inter-individual BOLDamplitude differences. The former are partially due to local differences in the signal-to-noise
ratio of the fMRI setup.
4.3 Results
Previous studies found that V1 is most responsive to (L-M)-cone stimulation, less
responsive to (L+M)-cone stimulation, and least responsive to pure S-cone stimuli using the
cone-contrast metric (Engel et al., 1997b; Liu & Wandell, 2005). We asked if this holds true
for extra-striate visual areas using high constant cone-contrast (14%), and extended the
stimulus range with additional cone-contrast-space-directions, such as S-(L+M) and L+M+S.
63
EXPERIMENT 2- Speedtuning
Results
We also asked if known sensitivity differences between chromatic responses within a given
visual area change when the stimulus pattern velocity is varied at 1, 4, or 10Hz. Finally we
aimed to find response pattern differences as predicted by behaviorally identified motion
channels (Gegenfurtner & Hawken, 1996)
Figure 4.2 shows examples for the localization of the V1/V3A/V4 and MT+ ROIs.
Figure 4.3 shows the normalized BOLD responses in the four visual areas investigated,
averaged over subjects.
A
B
Ring stimulation
C
Wedge stimulation
D
3D-rendering & ROIs overlay
E
F
Wedge stimulation Ring stimulation
V3A
V1
hV4
MT+
V3A
V1
hV4
V3A V1
hV4
LEFT
RIGHT
hV4
V1 V3A
hV4
V1 V3A
G
H
V3A
LEFT
RIGHT
V3A
V3A
V1
hV4
MT+
V1
hV4
MT+
MT+
hV4
V1
3.86 5.62 t-value 9.13 10.88
Figure 4.2: Borders of visual areas and location of region of interest.
V1 (blue), V2 (red), V3 (yellow), V3A (magenta) and hV4 (green) on cortical flat-maps determined by
expanding ring- (A,F) and rotating wedge-stimulation (B,D) for the left (A,B) and right hemisphere
(E,F) of one subject. MT+ (cyan) was identified using an independent localizer scan contrasting
moving vs. stationary dot-fields (see experiment 1). 3D-renderings of the left (C) and right (D)
hemispheres of the same subject with the visual areas overlaid (lateral view).
Panels G and H describe the identification of regions of interest (ROIs) for the left and right
hemisphere respectively. SPM99 was used to determine where any of the stimuli elicited a significant
activation in the occipital lobe (contrast stimulus vs. gray background). The resulting t-image and the
pre-determined borders of visual areas were overlaid onto flat maps to identify activity clusters within
the visual areas. The data from a cubic ROI (27 voxels) around this peak voxel were used in the
subsequent analysis. The colored boxes show the contrast t-image, overlaid on T1-weighted slices of
the same subject. The circles indicate the approximate location of the center-peak-voxel within V1
(blue-box), hV4 (green-box), MT+ (cyan-box) and V3A (magenta-box).
64
EXPERIMENT 2- Speedtuning
Results
4.3.1 Analysis of variance
4.3.1.1 Global BOLD-Analysis
The means (over twelve brain hemispheres) of the standard deviation normalized
BOLD-responses obtained under all stimulus conditions in the four visual areas under
investigation are shown in Figure 4.3. A three-way repeated measures analysis of variance
(rANOVA) revealed significant main effects of brain area (four levels: V1,V3A, hV4, MT+
F(3,33)=7.74, p<0.001), of color (five levels: L+M, L-M, S-L-M,L+M+S and pure S,
F(4,44)=41.72, p<0.001), and a significant interaction between these two factors
(F(12,132)=2.37, p<0.01). Interstingly, there was only a marginal effect of pattern velocity
(three levels: 1Hz, 4Hz, 10Hz, F(2,22)=2.73, p<0.088) in this global analysis but the
interactions between velocity and both color or brain area were significant (velocity-color:
F=(8,88), p<0.001; velocity-brain area:
F(6,66)=4.0, p<0.005). Finally, a three way
interaction (F(24,264)=3.88, p<0.001) indicates that the investigated brain areas are
differently activated by the color-velocity combinations used in the experiment. In the
following these interactions will be further explored to investigate how color and velocity
interact in the different brain areas.
4.3.1.2 Analysis by velocity
This analysis investigates whether the relative amplitudes of the BOLD-responses
elicited by the different colors are similar in the four brain areas over velocity. The analysis
can also reveal possible changes in the preference for specific colors in the investigated brain
areas over speed. Three two-way ANOVAs were computed with factors color and brain area
(all five colors and all four brain areas), one for each velocity. Color has a highly significant
effect at all speeds (1Hz: F(4,44)=23.13; p<0.001; 4Hz: F(4,44)=40.3, p<0.001; 10Hz:
F(4,44)=11.0, p<0.001) indicating higher BOLD-responses for some colors than for others at
all speeds.
Note that the effects on the factor brain area have no clear interpretation when BOLDresponses are analyzed. Conversely, the interaction between brain area and another factor has
a clear interpretation and is informative on how different brain areas process different stimuli.
However, the interaction between color and brain area reaches significance only at the highest
velocity presented (1Hz: F(12,132)=1.5, p>0.1; 4Hz: F(12,132)=1.8, p>0.05; 10Hz:
F(12,132)=5,1, p<0.001) indicating that the brain areas respond similarly to stimuli with
different chromatic content at the slower velocities (1Hz and 4Hz), but differ at the highest
velocity (10 Hz). The qualitatively same results were obtained when only responses to stimuli
65
EXPERIMENT 2- Speedtuning
Results
that emphasize cardinal mechanism were included in the analysis (L-M, L+M+S, and S). The
results indicate that the BOLD-responses elicited by different chromatic stimuli differ
between brain areas, but mostly at higher velocities.
4.3.1.3 Analysis by brain area
Separate two-way ANOVAs for each brain area were computed, aiming to reveal
effects of stimulus combinations of speed and color within each visual area. In V1 we find a
main effect of color (F(4,44)=20.1; p<0.001) but no main effect of velocity (F(2,22)=0.99;
p>0.25), and a significant interaction between color and motion was found. A similar pattern
is found in area MT. While color has a strong differential effect on the BOLD-activation in
this area (F(4,44)=27.53; p<0.001) no differential response over speeds was found
(F(2,22)=0.1; p>0.5) but, as in V1, a significant interaction (F(8,88)=4.6; p<0.001).
Areas V4 and V3A provide a somewhat different picture. Similar to V1 and MT both
show an effect of color (V4: F(4,44)=26.7, p<0.001; V3A: F(4,44)=16.9, p<0.001). In contrast
to V1/MT areas hV4 and V3A show an additional effect of velocity (V4: velocity
F(2,22)=9.2, p<0.001, interaction: F(8,88)= 7.96, p<0.001; V3A: velocity F(2,22)=4.5,
p<0.05). These results suggest a greater similarity in the processing of V1/MT as compared to
V3A/V4. Figure 4.3 suggest that the differential velocity effect is due to the fall-off of the
BOLD-responses in areas V3A and hV4 as temporal frequency increases. Conversely, areas
V1 and MT+ do not show such an effect. V1 and MT+ revealed a response profile one might
expect from brain areas that represent velocity in populations of neurons tuned to a wide range
of velocities. However, the significant color-velocity interactions we found in all visual areas
indicate that the main effects of velocity might differ between different color-directions.
Therefore, we analyzed the effects of velocity for each color separately.
66
EXPERIMENT 2- Speedtuning
Results
Figure 4.3: Speed tuning in V1, MT+, hV4 and V3A for five cone-contrast
directions and three speeds
Abscissa: cone-contrast colors; Ordinate: amplitude of standard deviation scaled response; Bars:
yellow=1Hz, blue=4Hz, red=10Hz; hV4 and V3A exhibit response decline as temporal frequencies
reach 10 Hz for achromatic (L+M+S), luminance (L+M), B/Y (S-(L+M)) and R/G (L-M) stimuli indicative
of low-pass properties. MT+ responses show slight increases at 4 Hz responses indicative of bandpass properties. V1 responses increase for L-M as temporal frequency is raised. Low S-cone
responses as compared to extra-striate regions.
4.3.1.4 Velocity effects by cone-contrast direction
The effects of the stimuli that emphasize the putative detection mechanism responses
(L-M, L+M+S, S) were analyzed separately for each visual area investigated. The critical Fvalue for the Bonferroni-correction of the 12 comparisons (3 cone-contrast directions in 4
brain areas) is 7.1 and the corresponding p-value is 0.0042. For completeness the uncorrected
p- and their corresponding F-values are report.
The BOLD-response to the L-M stimulus changed only in V1 significantly over the
three speeds (V1: F(2,22)=7.89, p<0.003). In V3A, hV4 and MT+ there appears to exist a
non-significant tendency that the 4Hz L-M stimulus elicits the highest response (see Figure
4.3; V3A:F(2,22)=1.8, p>0.1; hV4: F(2,22)=1.7, p>0.1; MT+: F(2,22)=2.56, p>0.1). The
strongest effect is found in the decrease of the L+M+S response with increasing speed in areas
V3A (F(2,22)=11.0, p<0.001) and V4 (F(2,22)=17.24, p<0.001). Neither in V1 (F(2,22)=4.1,
p<0.05) nor in MT (F(2,22)=4.78, p<0.025) was the decrease significant when corrected for
multiple comparisons, but the effects were of similar size in these two areas. The apparent
increase with increasing velocity of the S response reaches in none of the brain areas
67
EXPERIMENT 2- Speedtuning
Results
significance. However, the effect sizes observed draw a similar picture as for the L+M+S
stimulus. V1 and MT+ show the least effect sizes (V1: F(2,22)=0.24; MT+: F(2,22)=2.58,
both p>0.05) and both would not reach significance, even without correction for multiple
comparisons. In areas V3A and hV4 stronger effect sizes are obtained that would reach
significance without the correction (V3A: F(2,22)=4.53, p<0.025) and hV4 (F(2,22)=6.78,
p<0.005). The most important source for the interactions by velocity found in the global
ANOVA appears to be the weaker response to the L+M+S stimulus with increasing speeds in
areas V3A and hV4. Since the L+M+S stimulus emphasizes the response in the luminancemechanism this suggests that the effect is due to the luminance contrast. This could be
interpreted as an indicator that V3A and hV4 exhibit the properties expected from the slow
mechanisms for motion perception since regions associated with the fast mechanisms should
not exhibit substantial BOLD-response decline as the velocity of luminance pattern increase
(Gegenfurtner & Hawken, 1996).
4.3.2 Cluster analysis
One tendency suggested by the ANOVAs was that BOLD responses obtained at the
various color and velocity combinations are more similar between areas V1 and MT and more
alike in areas V3A and hV4. A hierarchical cluster analysis on the responses to the fifteen
color-velocity combinations obtained in each hemisphere was performed to investigate the
similarity between the response profiles of the four visual areas.
Areas V3A and hV4 cluster more closely than any other group. The correlations
between the mean response vectors in the 15 conditions investigated (3 velocities and 5 color
directions) are listed in Table 4.1. The Pearson-correlation coefficient measures the angle
between two response vectors. This angle can be interpreted as a measure of similarity
between the response profiles of two visual areas and neglects differences in BOLD-response
strength. As already indicated by the cluster analysis, the highest similarity was found
between V3A and hV4 (r=0.974). This correlation is higher than between any other pair of
visual areas (at least p<0.05 for all comparisons). The angle between the feature vectors of
V3A and hV4 is nearly a third of the angle between the other pairs, specifically V3A and
MT+.
68
EXPERIMENT 2- Speedtuning
V1
V1
1
V3A
33.02
hV4
30.44
MT+
30.78
Angle (deg)
V3A
0.838
1
13.05
34.29
Results
Correlation
hV4
MT+
0.862134 0.859132
0.974161 0.826175
1
0.81027
35.88
1
Table 4.1: Results of cluster analysis and Pearson-correlation coefficient
Response profiles of hV4 and V3A are highly similar as indicated by the highest correlation coefficient
and lowest angle difference of the response vectors of these two regions compared to any other pair.
4.3.2.1 Stimuli for best separating hV4 and MT+
Color and velocity are stimulus attributes that are often used to separate dorsal and
ventral stream processing. We performed an exploratory analysis to find the best separating
combinations. Therefore we calculated the differential effects (t-values) for all stimulus
combinations in areas hV4 and MT+ (see table 4.2 in the appendix). These areas are often
regarded as representative for the ventral and dorsal stream, respectively (Livingstone &
Hubel, 1988). Our stimulus set comprises 105 unique combinations of color and motion.
The separation of the ventral and dorsal stream could be achieved by at least two
approaches: Stimulus combinations that modulate one stream without having a significant
effect in the other could be used to modulate one stream selectively. Furthermore, stimulus
combinations that produce a strong positive effect in one stream and a reverse effect in the
other stream could be used to emphasize one stream and simultaneously suppress the other.
As for the first approach we found that the comparisons (L-M 4Hz) – (L+M+S 10Hz)
and (L+M 1Hz) – (S 1Hz) produced highly significant effects in hV4 but no effects in MT+,
after Bonferroni-correction for multiple comparisons. In these comparisons the effects in hV4
exceed those in MT+ by more than seven t-values (for a complete list of corrected t-values see
appendix table 4.3). Area MT+, but not area V4 exhibits the most reliable effects in the
comparisons (L+M 10Hz) – (S 4Hz), (L+M 10Hz) – (S 10Hz), and (L-M 10Hz) - (S 10Hz). In
these comparisons the effects in MT+ exceed those in hV4 by more than 6 t-values. It is
interesting to note that most stimulus combinations that favour effects in hV4 involve
combinations of L+M, L-M or S-L-M equal to or below 4Hz as the minuend and L+M+S >=
4Hz or S at 1Hz as the subtrahend, i.e. the first lead to higher BOLD-responses in hV4 than
the second. These results match the general picture of hV4 as an area that prefers slow
69
EXPERIMENT 2- Speedtuning
Discussion
chromatic over fast achromatic stimuli. However, it should be noted that area MT+ shows no
preference for the above combinations, in neither direction of the subtraction. This argues
strongly against area MT as an area that generally prefers fast achromatic over slow chromatic
stimuli. The full set of results is given in table 4.3 in the Appendix.
As for the second approach, we found no stimulus combination that produced a
significant positive effect in one area and a concurrent negative effect in the other. All
differences between conditions that were significant in both MT+ and hV4 went in the same
direction. However, among these stimulus combinations the largest differences between the
effects in hV4 and MT were found in the comparison (L+M 10Hz) – (L+M+S 10Hz), with an
effect in V4 that was by 4.5 t-values higher than in MT+. The strongest effect in favour of MT
was found for the combination (L+M 4Hz) – (S 4Hz). Here the MT effect exceeded the hV4
effect by 4.14 t-values. Such effects were only found with stimulus combinations involving Sstimuli (see table 4.4 in the appendix).
4.4 Discussion
The goal of this experiment was to investigate response differences depending on
stimulus color and velocity at one fixed cone-contrast level in several visual areas known to
be involved in cortical color and/or motion processing.
On the basis of differential response patterns in several visual areas to various stimulus
color and speed combinations, associations regarding psychophysically determined motion
perception channels were achieved. It was shown, that BOLD-responses to equal conecontrast stimuli differ between brain areas, but mostly at higher velocities. Additionally, best
stimulus combinations to separate the ventral (hV4) and dorsal (MT+) path were presented.
4.4.1 Two motion pathways in human cortex
Gegenfurtner et al. (1996) proposed two non-color-blind processing streams for
motion perception that differ mostly in their temporal characteristics (Gegenfurtner &
Hawken, 1996). The authors describe a fast motion pathway, that veridically represents the
velocity of moving patterns, and also responds to moving isoluminant stimuli, but might not
code the color of such patterns. A putative second slow pathway is suggested to show a high
sensitivity to color, signals the direction of slowly moving patterns, but its coding of stimulus
velocity is not veridical. The authors attributed visual area MT as a likely candidate for the
fast motion pathway and hypothesize area V4 to be involved in the slow motion mechanism
(Gegenfurtner & Hawken, 1996).
70
EXPERIMENT 2- Speedtuning
Discussion
Visual areas that correspond to the fast motion perception mechanism are expected to
respond invariantly to the stimulus velocity and also show responses to chromatically defined
moving patterns. On the other hand, cortical regions that correspond to the slow motion
perception mechanism are expected to show high sensitivity to chromatically defined moving
patterns especially at low speeds, while responses should decline as velocities increase above
4Hz (Gegenfurtner & Hawken, 1996).
In concordance with this view our measurements show a relative invariance of MT+
responses to temporal frequency regardless of color direction while hV4 and even more so
area V3A, show a clear response decline at high temporal frequencies.
This effect was most obvious for achromatic moving stimuli, which emphasize a proposed
luminance detection mechanism (Derrington & Henning, 1993; Krauskopf, 2000a). Responses
in visual area hV4 and V3A to achromatic moving gratings declined substantially, while no
such effect was observed for visual areas MT+. In addition all three regions respond well to
moving chromatic stimuli, especially area hV4 and V3A, for which BOLD-response profiles
to chromatic gratings, suggest a preference for slowly moving R/G-gratings up until 4Hz.
We thus suggest neurophysiological evidence for a color-motion processing network
derived psychophysically and set forth by Gegenfurtner et al., in which MT+ responds to all
temporal frequencies regardless of color direction and thus accounts for a fast motion
mechanism, while V3A and hV4 seem to prefer slowly moving chromatic stimuli as their
response declines dramatically as temporal frequencies exceed 4Hz and thus account for the
slow mechanism of motion (Gegenfurtner & Hawken, 1996).
Further evidence supporting several cortical motion processing sites comes from a
behavior study investigating chromatic detection and direction of motion thresholds in
individuals with cerebral achromatopsia using slow stimulus velocities (2Hz) and fairly high
chromatic contrasts (Cavanagh et al., 1998). This study found that the subjects, despite their
profound loss in the subjective experience of color and their inability to detect the motion of
faint colors, showed surprisingly strong responses to high-contrast, moving color stimuli,
equal in all respects to the performance of subjects with normal color vision (Cavanagh et al.,
1998). The authors could not determine the cortical site, which may be responsible for
processing slow moving isoluminant motion stimulation. They were able to dismiss ventral
region hV4 as a likely candidate, since it is severely damaged in cerebral achromatopsia
(Zeki, 1990). They additionally argued against area MT+ as likely candidate, because
chromatic sensitivity of monkey MT+ neurons is not sufficient to explain the excellent
chromatic motion sensitivity (Gegenfurtner et al., 1994; Cavanagh et al., 1998).
71
EXPERIMENT 2- Speedtuning
Discussion
Data presented here suggest area V3A to be responsible for nearly normal chromatic
motion thresholds in subjects suffering from cerebral achromatopsia at slow moving high
contrast color patterns. V3A is a likely candidate, because it exhibits a preference for slowly
moving chromatic stimuli. In addition it is located on the dorsal side of the human occipital
cortex, rendering it highly unlikely to be damaged in cerebral achromatopsia, which follows
damage to the ventral occipital cortex. Furthermore it is distinct from lateral occipital region
MT+ and represents a complete hemifield (see section 3.3.1 and 3.3.2). Finally there is
evidence in our data for color-opponent mechanisms in V3A, which is further discussed in the
next section.
4.4.2 Color opponency in visual cortex
A long lasting controversy regarding color-opponent mechanisms in visual cortex is
fueled by neurophysiological findings reporting relatively few color-opponent cells in V1
(Livingstone & Hubel, 1984), while an opposing view is supported by neuroimaging studies,
which suggest that such neurons might be relatively numerous (Schluppeck & Engel, 2002).
Schluppeck et al. (2002) argue that when cortical responses to L-M stimulation exceed those
for luminance, given equal cone-contrast, a color-opponent mechanism must be the
underlying source. Applying this criterion to our data set reveals strong color-opponent
responses not only in V1, but also in V3A and not surprisingly in hV4. In all of these visual
areas L-M responses exceeded luminance responses.
Color opponent mechanism in motion selective cortex (MT+) are controversial
because this area is believed to receive mainly magnocellular and thus luminance input, which
sum cone signals, as opposed to differentiating them in an opponent fashion (Maunsell et al.,
1990). The applied stimulus set in the present experiment contained two color directions that
emphasize the magnocellular pathway, namely L+M+S and L+M. If indeed MT+ solely
processes luminance information, responses to chromatically defined stimuli such as L-M,
emphasizing the color-opponent, parvocellular pathway, should be substantially lower in that
region compared to luminance generated signals. Our data do not show response patterns
supporting a clear luminance preference in MT+. Instead, we find, BOLD-response patterns,
elicited by stimuli emphasizing parvo- and magnocellular mechanism to be roughly equal.
This finding does not allow for strong conclusions regarding the presence of color-opponent
mechanisms in MT+, applying the criterion set forth by Schluppeck et al. (Schluppeck &
Engel, 2002). The results are however in concordance with and consolidate recent findings
from neurophysiological reports on monkeys and human imaging studies, showing that MT+
72
EXPERIMENT 2- Speedtuning
Discussion
may adhere to signals from both magno- and parvocellular pathways (Seidemann et al., 1999;
Wandell et al., 1999; Nassi et al., 2006).
Furthermore, above discussed study by Cavanagh et al. (1998) reported that
individuals with cerebral achromatopsia perceive moving isoluminant colored stimuli just as
well as normal observers, given a fairly high cone-contrast(Cavanagh et al., 1998). Cavanagh
et al. argue that cortical sites outside of MT+ must mediate chromatic motion signals based on
the assumption, that MT+ receives little or no parvocellular input. Another group, however,
showed clear ‘color signals in human motion-selective cortex’, albeit pure S-cone-signals,
arguing that the MT pathway contains enough information to permit individuals with cerebral
achromatopsia to perform motion judgement (Seidemann et al., 1999; Wandell et al., 1999).
Our data support both arguments. The strong L-M responses in MT+ suggest that it contains
enough parvocellular input to allow individuals with cerebral achromatopsia to perform
veridical chromatic motion judgments. Nonetheless, another motion sensitive area, V3A, is
also highly responsive to L-M stimulation, rendering it a candidate for chromatic motion
processing independent of MT+, as well as exhibiting properties that suggest color-opponent
mechanisms in this dorsal region.
4.4.3 Colortuning
Many neurophysiological studies on monkey visual cortex aim to determine the
specialization of single cells regarding a specific stimulus attribute, for instance color, by
systematically varying certain stimulus parameters, as for instance contrast or temporal
frequency (Gegenfurtner et al., 1994; Hawken et al., 1996; Seidemann et al., 1999; Solomon
& Lennie, 2005; Conway & Livingstone, 2006). An indication for specialization of a given
neuron in these studies is typically a higher firing rate to one stimulus configuration as
compared to another. The stimulus configuration yielding the highest firing rate is associated
with the cells stimulus preference, while the response profile over many stimulus attribute
levels is called the tuning with respect to the investigated stimulus attribute.
Recent human imaging studies have adopted this approach to determine the specific
tuning properties of cell conglomerates in specific regions of the human brain or
retinotopically mapped visual field representations (Engel et al., 1997b; Liu & Wandell,
2005). An indication for stimulus preference in fMRI-studies is typically the magnitude of the
measured BOLD-response. Higher BOLD-responses elicited by a certain stimulus level A
compared to lower BOLD-responses elicited by stimulus level B are interpreted as a
preference to stimulus level A of the region yielding the response differences. It is thus
73
EXPERIMENT 2- Speedtuning
Discussion
possible to determine the tuning of a human visual area with respect to specific stimulus
attributes.
To further advance the understanding of human cortical color and motion processing
we aimed to measure speed tuning curves using equal cone-contrast colors which will be
discussed in the next section. In this section the characterization of the visual areas with
respect to the measured BOLD–responses depending on different chromatic stimulation is
discussed. By comparing BOLD-responses elicited by stimuli at equal cone-contrast levels of
14% we can draw conclusions regarding the colortuning of the measured visual areas for this
specific contrast level. It has to be noted that the drawn conclusions regarding colortuning
may only be valid at the used cone-contrast level and one should be cautious in extrapolating
the results to other cone-contrast stages because the relative response strengths (the tuning)
may change when cone-contrast is varied. Extensive color and speed measurements at
different chromatic contrast levels in experiment 3 of this thesis will investigate this
possibility.
For the visual areas V1, hV4 and V3A, we observed a response magnitude hierarchy in
the order of L-M > S-(L+M) > L+M > L+M+S > S. This response hierarchy can be
interpreted as color-tuning property for these areas at equal 14% cone-contrast. The results for
V1 are well in line with a prior imaging study in humans, reporting greater sensitivity in
primary visual cortex to L-M stimuli compared to L+M stimulation (Engel et al., 1997b). As
an extension to this prior V1 report, our results suggest that the colortuning properties at 14%
cone-contrast do not change substantially in extra-striate regions with the exception of MT+,
which responds equally to all but S-cone stimulation. BOLD-responses in human MT+
emphasizing solely S-cone contribution is small but present. This observation is in
concordance with an fMRI study showing markedly reduced sensitivity of neurons in MT+ to
S-cone stimuli compared to luminance (L+M) stimulation (Wandell et al., 1999). Wandell et
al. (1999) also reported higher responses in V1 to L+M as compared to S-cone stimuli, which
is in concordance with the measured response hierarchy described above (Wandell et al.,
1999).
The underlying source for human cortical color responses is still an open issue.
Neurophysiological monkey studies have suggested some links to provide insight into the
neuronal sources of the observed BOLD signals in humans. A recent study was able to
convincingly show that signals stemming from the three classes of neurons, organized in three
layers of the lateral geniculus (see section 1.2.3.1), are transmitted via distinct thalamocortical pathways and terminate into separate projection sites in different layers of primary
visual cortex (Chatterjee & Callaway, 2003). The spatial resolution of fMRI is currently not
74
EXPERIMENT 2- Speedtuning
Discussion
able to resolve distinct cortical layers. This restriction renders application of
neurophysiological findings in monkey cortex to human cortex speculative because it cannot
be tested with the same methods. However, after methodological progress has occurred over
the past years, many findings from the monkey model have subsequently been confirmed in
humans (Dougherty et al., 1999; Seidemann et al., 1999; Wandell et al., 1999), and hence
motivate a discussion of the human colortuning properties found in experiment 2, in the realm
of properties of single cells in monkey visual cortex.
Our data obtained for primary visual cortex suggests a preference for R/G over B/Y
responses, which may either indicate a higher sensitivity of color selective cells to R/G or a
higher neuronal density for R/G preferring cell populations. Several groups have found that
there are more red-green double-opponent cells than blue-yellow ones, supporting the later
argument and suggesting double-opponent cells as a candidate neural source of the chromatic
tuning properties in our data (Wiesel & Hubel, 1966; Livingstone & Hubel, 1984; Ts'o &
Gilbert, 1988; Conway & Livingstone, 2003). Some groups have confirmed the existence of
double-opponent cells in primary visual cortex which are sensitive to color and luminance
contrast (Thorell et al., 1984; Conway, 2001; Johnson et al., 2001; Conway et al., 2002;
Johnson et al., 2004). It is thus also likely that observed responses to R/G and luminance
stimuli were generated by the same neural population of cells.
There have been speculations whether distinct cell types may be responsive to more
than just one stimulus attribute (Gegenfurtner et al., 1996; Gegenfurtner & Kiper, 2003), and
that for instance double opponent color-luminance cells might multiplex color, form (Johnson
et al., 2004) and even direction of motion (Conway & Livingstone, 2003). The applied
methodology does not allow for strong conclusion regarding multiplexing of several stimulus
attributes other than to note that temporal frequency of the moving patterns does have an
effect on the measured subpopulation of neurons responding to our stimulus configurations,
which will be discussed in the next section and further investigated in experiment 3.
However, neurophysiological studies regarding direct projections from V1-colorsensitive-blobs to the form ‘sensitive’ area V4 suggest, that color selective cells in V1 might
be suitable for tracking moving isoluminant borders (Schein & Desimone, 1990; Ghose &
Ts'o, 1997; Pasupathy & Connor, 2001). V1 may thus build the initial encoding stage for
signals of moving objects, which might further be integrated in hV4 where the receptive field
size is much higher (Zeki, 1978a; Smith et al., 2001) and neurons are also direction selective
(Nealey et al., 1991; Tolias et al., 2001b).
75
EXPERIMENT 2- Speedtuning
Discussion
4.4.4 Speedtuning
As described in the previous section, the response profile over several stimulus levels
for one visual attribute is called the tuning with respect to the investigated stimulus property.
When more than two stimulus attribute levels are measured the response pattern can be
interpreted as tuning curve for this specific stimulus attribute. Experiment 2 applied three
different rotation velocities to every chromatic stimulus patterns. It was thus possible to derive
speedtuning curves per color direction and visual area.
The speed tuning curves for a given color-direction are similar across all extra-striate
cortical areas measured (hV4, MT+, V3A), while marginal deviations from that scheme are
observed in striate cortex (V1). This deviation is especially prevalent for high-frequency L-M
modulated targets compared to luminance targets, indicating differences in temporal filter
properties between striate and extra-striate cortex.
The first fMRI-study applying retinotopic mapping as well as parametric modulation
of cone-contrast presented iso-response curves of BOLD-signal responses in area V1 and V2
to three temporal frequencies of a cone-contrast reversing checkerboard flicker stimulation
(Engel et al., 1997b). Their data agree well with a prior study (Kleinschmidt et al., 1996)
showing highest responses in striate cortex to L-M-stimulation, which we confirm as well
here (see section colortuning).
However, the temporal frequency measure used by Engel et al., cannot be regarded as
real motion stimulation since they used contrast reversing checkerboard flicker which has
recently been shown to activate substantially less than real motion stimulation in area V1 and
MT+ (Liu et al., 2004; Muckli et al., 2005). Our V1 L-M responses reveal higher responses to
higher temporal frequencies which differs from data presented by Engel et al. (Engel et al.,
1997b). The divergence may arise because of stimulation differences in the temporal domain
as well as substantially higher equal cone-contrast in our study. The applied stimulation in the
present experiment did move (rotate) in the visual field, simulating real motion, which is not
the case for stationary flicker stimuli. However, it is long known that some color-opponent
neurons in primary visual cortex can follow high temporal frequencies well above
heterochromatic fusion frequencies (Gur & Snodderly, 1997). It may thus be possible to find
strong responses to color-opponent stimulation at high temporal frequencies in V1. But why
do we observe this increase only in V1 for one color direction?
Since the 14% cone-contrast was constant in all color-directions, the L-M stimulation
was the most salient. Although every effort was made to keep the attentional load constant
and away from the rotating color patterns by means of a demanding number discrimination
task at the fixation cross, we cannot completely rule out that automatic attentional
76
EXPERIMENT 2- Speedtuning
Discussion
mechanisms attributed to L-M responses at increasing velocities. It has been shown that popout stimulus context effects, such as a highly salient, 14% equal cone-contrast pattern, might
bias visual processing and this effect seems to have its origin in early visual cortex (Beck &
Kastner, 2005; Yantis, 2005). In addition it has been shown that color and luminance
stimulation attract independent attention (Morrone et al., 2002). Moreover, there is recent
evidence that a change-detection mechanism sensitive to unattended changes in motion
direction may exist in the human visual system (Pazo-Alvarez et al., 2004).
A recent study reported the first parametric fMRI-measurements in which conecontrast and the stimulus velocity were systematically modulated (Liu & Wandell, 2005).
Most of the results agree well with the present study such as color-opponency in striate- and
extra-striate cortex and its implications (see section 4.4.2 color-opponency) but some of the
inferences need to be reconsidered in the realm of this thesis.
Liu and Wandell propose a common functional network between area V3A and MT+
by referring to two studies that found common response profiles to achromatic stimuli
(Tootell et al., 1997; Smith et al., 1998). These two studies have only used achromatic
stimuli. If L-M stimulation would have been tested as well, they may have found differences
between V3A and MT+ at least at higher frequencies. In addition Liu and Wandell state that
V3A and MT+ have weak responsivities to L-M and pure S-cone stimuli compared to other
visual areas as well as profoundly different properties than a ventral occipital area. Our data,
and partially theirs, suggest otherwise. We find equal responsivity in MT+ to L-M and L+M
stimuli regardless of temporal frequency as well as higher activity in V3A to L-M than to
L+M stimuli which both decrease at frequencies higher then 4Hz in this region. Also, the hV4
L+M-speedtuning curve is quite similar to the V3A L+M-speedtuning.
We speculate that the similarities in the measured response profile between V3A and
hV4 might be indicative of extensive cross talk between these areas in a common network that
interprets and integrates motion signals, may they be of chromatic or luminance nature. Our
data suggests a functional color-motion-network comprised of at least hV4, MT+ and V3A
(leaving open additional contribution from future motion areas), which responds well to all
chromatic modulations.
The equal cone-contrast level of 14% differs substantially from prior investigations,
characterizing chromatic response properties in human cortex (Kleinschmidt et al., 1996;
Engel et al., 1997b; Liu & Wandell, 2005) and may partially be responsible for some
diverging conclusions (Liu & Wandell, 2005). To test the effects of different chromatic
contrast levels, the following experiment will provide extensive fMRI measurements using
77
EXPERIMENT 2- Speedtuning
Discussion
visual stimuli that besides differing in chromatic axes and velocity will be modulated in the
contrast domain.
78
EXPERIMENT 3 - Colortuning
Introduction
5 EXPERIMENT 3 - Colortuning
5.1 Introduction
Experiment 2 investigated BOLD-response profiles in several visual areas to various
combinations of stimulus color and velocity at an equal cone-contrast level. The measured
response profiles were used to associate visual areas to one of two mechanisms of motion
perception, because specific properties and limitations of these mechanisms have been
characterized neurophysiologically and psychophysically (Gegenfurtner et al., 1994;
Gegenfurtner & Hawken, 1995, 1996). These reports provided predictions with respect to the
expected temporal characteristics and chromatic sensitivity of visual areas associated with
either of the two motion channels. On the basis of these predictions we generated two criteria.
The first criterion pertains to the temporal properties of the two channels and was tested in
experiment 2. It states that one mechanism is not sensitive to stimuli moving at high temporal
frequencies, rendering it the slow motion channel, while the second is responsive to stimuli
moving at a wide range of temporal frequencies, hence called the fast motion channel.
The data obtained in experiment 2 suggested an association of area MT+ to the fast
motion channel, because high temporal frequency stimulation did not reduce responses
compared to slow or medium temporal frequencies in that region. Contrarily, V3A and hV4
responses to high temporal frequency stimuli declined substantially while responding
considerably well to low and medium temporal frequencies. On the basis of the observed
response characteristics of these two areas, an association with a perceptual slow motion
channel was deduced. The first criterion to assign visual areas to either of two
psychophysically derived motion channels, differing in their temporal characteristics, was
supplied.
The second criterion to associate visual areas to the proposed motion channels pertains
to differential chromatic contrast sensitivities of the two channels and will be part of the
investigations in experiment 3. This criterion is directly related to the original motivation to
employ a two-channel model as opposed to only one channel for motion perception, because
no single region was neurophysiologically identified, that could account for luminance as well
as chromatic motion thresholds observed in monkeys and humans (Metha et al., 1994;
Gegenfurtner & Hawken, 1995; Stromeyer et al., 1995). For instance, the candidate region
MT+, known to be involved in motion perception (Newsome et al., 1985; Newsome et al.,
1991) could not be held responsible for the excellent human detection and discrimination
sensitivity for slowly moving red-green stimuli, because MT+ neurons in the macaque
monkey did not respond sufficiently well to low contrast chromatic patterns (Metha et al.,
79
EXPERIMENT 3 - Colortuning
Introduction
1994; Gegenfurtner & Hawken, 1995; Stromeyer et al., 1995). To be specific, a trained
monkey can detect drifting, chromatic gratings with color contrasts lower by a factor of three,
than can be detected by even the most sensitive MT cells (Gegenfurtner et al., 1994). Visual
areas responding to low contrast moving chromatic stimuli other than MT+ must therefore
exist and play some role in the detection and discrimination of motion defined by colorcontrast edges (Thiele et al., 1999). The second criterion to associate visual areas to either of
the two motion channels must thus require the candidate reasons to exhibit a higher sensitivity
to chromatic contrasts than area MT+, in order to account for human detection and
discrimination sensitivity specifically to slowly moving red-green stimuli for which
behavioral data is established (Metha et al., 1994; Gegenfurtner & Hawken, 1995; Stromeyer
et al., 1995).
As data from experiment 2 has shown, visual areas hV4 and V3A seem to prefer slowly
moving chromatic stimuli up to 4Hz at equal cone-contrast of 14%. However, chromatic
contrast sensitivity cannot be judged from experiment 2, because 14% cone-contrast is well
above human detection threshold (Mullen, 1985) and all visual areas responded strongly to
these stimuli. To measure cone-contrast sensitivity in a cortical region, several cone-contrast
levels have to be measured. The optimal contrast levels should range from contrasts at which
the region does not respond to contrasts at which the region responds maximally. Using fMRI
in humans, which measures neural population responses, the appropriate cone-contrast levels
for sensitivity measures of visual areas cannot directly be inferred from single-cell threshold
data. It is thus appropriate to make a choice of contrasts, which contain low contrast values for
each cone-contrast directions of interest, in order to uncover response sensitivity differences
between candidate areas for the neural correlates of the slow motion channel. Furthermore the
cone-contrast stages should also contain sufficiently high contrasts to allow comparisons with
Experiment 2. If higher chromatic contrast sensitivity in hV4 and V3A as compared to MT+ is
found, the second criterion, stating higher chromatic contrast sensitivity in regions associated
with the slow motion perception channel would be fulfilled and the proposed associations of
cortical regions to the two motion channels, as suggested in Experiment 2 could be
consolidated. Experiment 3 will employ 4 cone-contrast stage values along each color
direction under study ranging from 3% to 100% of the maximally possible cone-contrast,
which may differ for the respective color directions (see section 2.1.1.1).
The chromatic stimulus conditions used in experiment 2 were chosen to lie along
directions in cone-contrast-space emphasizing post receptoral mechanisms (see section
2.1.1.1). However, little is known about human cortical response characteristics to chromatic
patterns, stimulating along cone-contrast space directions other than the cardinal directions.
80
EXPERIMENT 3 - Colortuning
Introduction
An influential study by Engel et al. (1997) reported measurements in human subjects, using
contrast reversing checkerboard patterns modulated in two cone-contrast planes, in which the
putative chromatically opponent and achromatic mechanisms are located (Engel et al.,
1997b). Isoresponse contours were reported for V1 and V2.
In order to derive a more complete picture of human cortical chromatic response
properties, measurements along color directions, located in planes of cone-contrast space not
investigated yet, are thus of high interest. Furthermore, extra-striate regions are likely to
contribute to chromatic motion processing, as shown in experiment 2 and earlier reports
(Gegenfurtner & Hawken, 1996; Cavanagh et al., 1998; Thiele et al., 1999; Wandell et al.,
1999; Liu & Wandell, 2005). Experiment 3 will present results consistent with measurements
made by Engel et al. (1997) but extend the findings to visual areas hV4, V3A and MT+, in
addition to a more complete sampling of cone-contrast space. The chromatic stimulus range
will be extended to five planes in cone-contrast space including the two planes used by Engel
et al. to facilitate comparison with this influential report (Engel et al., 1997b).
As indicated above chromatic stimulation in each of thirteen color directions forming
the five planes will be parameterized in four cone-contrast stages to derive cone-contrast
response curves. These contrast response curves can be compared across retinotopicallymapped regions of visual cortex to derive color direction specific sensitivity differences. To
facilitate comparison with experiment 2, and further investigate the effect of stimulus
velocity, all parameterized color conditions will be presented at high and low stimulus
velocities.
Only two studies exist so far that combine parametric cone-contrast stimulation
control with stimulus modulation in terms of temporal frequency (Engel et al., 1997b; Liu &
Wandell, 2005). While the first, Engel et al. (1997), modulated the temporal frequency using
stationary contrast reversing cone-contrast checkerboard patterns, the second study, by Liu
and Wandell (2005), applied moving patterns, stimulating along only three axes in conecontrast space (Liu & Wandell, 2005). Liu and Wandell (2005) report cone-contrast response
functions for these three color conditions and two velocities in striate and extra-striate visual
areas (Liu & Wandell, 2005).
Unfortunately the later study had several shortcomings (Liu & Wandell, 2005).
Adaptational or sequential effects were not controlled, because each functional scan contained
only one cone-contrast condition and the cone-contrast levels were chosen in a way, which
made the comparison between color directions difficult. Furthermore the authors admit that
they could not obtain good signals from area V3A and MT+, which constitute cortical regions
of great interest in this thesis.
81
EXPERIMENT 3 - Colortuning
Methods
Results reported here aimed to overcome these shortcomings by extensively measuring
two experienced observers in a total of 18 fMRI-scanning sessions at 18 different days, while
presenting them with all color-contrast conditions in one session, using an event-related
stimulus presentation. At nine of these days the stimulus presentation protocol was conducted
with stimulus patterns moving at a velocity of 1Hz, while nine remaining measurements
presented the same stimulus presentation protocol at a temporal frequency of 10Hz (see
Methods, section 5.2). These extensive measurements yield reliable signals in every
retinotopically mapped visual area (see Experiment 1) and allowed for strong conclusions
regarding chromatic color contrast preferences throughout visual cortex and its dependence on
speed. Furthermore, the data provide a quantification of differential visual signal
representation, as the chromatic information is being passed on and processed from lower to
higher tier visual cortical areas in human subjects.
5.2 Methods
5.2.1 Subjects
Two male, right-handed subjects participated in the experiment. They gave their
written informed consent before scanning. The University of Magdeburg ethics committee
approved the experimental protocol.
5.2.2 Stimulus & experimental design
Stimuli were rotating rings with radial gratings presented in an annulus with 8.8°
visual angle outer radius and 2° visual angle inner radius (Figure 5.1). Two velocities
resulting in temporal frequency flicker of 1Hz and 10Hz were used in separate scanning
sessions. Direction of rotation was randomly assigned and changed on average every 4s. Each
color condition was presented at 4 different contrast levels of the maximally attainable RMScone-contrast per color direction (3%, 10%, 30% and 100%). For example in the L+M
direction, the maximal RMS-contrast was 96% therefore L+M stimuli were presented at 2.9%,
9.6%, 29% and 96% RMS-cone-contrast (see section 2.1.2). Stimulus colors were generated
along thirteen cone-contrast color-directions. Figure 5.2 illustrates the maximally attainable
cone-contrast colors for each color direction and the resultant percepts as different conecontrast values are applied. The background color was set to gray with the mean luminance
level of the grating (417 cd/m2).
The gratings were presented for 1s, followed by mean gray background without any
pattern presentation. The inter-stimulus interval (ISI) varied between 2s and 6s with a mean
82
EXPERIMENT 3 - Colortuning
Methods
ISI of 2.5s (mean gray presentation). The pseudo randomized sequence of ISIs and the
stimulus levels were optimized to allow for the efficient deconvolution of the event related
BOLD-response (Hinrichs et al., 2000).
Each run took 240s and ended with the presentation of a blank screen for 55s. Each
subject was scanned nine times (18 runs) for each of the two velocity condition. Over all runs
every stimulus combination of cone-contrast direction and cone-contrast level was presented
396 times. During the entire run subjects performed a demanding form discrimination task at
the fixation cross (tall cross vs. wide cross) with the aim to keep attention allocation constant.
The stimuli were back-projected onto a screen mounted in the magnet bore at 27cm distance
to the eyes and viewed via a mirror. The RGB-value to luminance functions were measured
(PR650 spectroradiometer, SpectraScan) for all three colors of the LCD-projector (Sharp,
SX21) and linearized with a lookup table (see section 2.1.2).
Figure 5.1. Representative stimulus
pattern used in Experiment 3. Adjacent
wedges contain colors stimulating along one
direction in cone-contrast space. This example
represents colors modulated along the L+Mdirection (96% RMS-cone-contrast, equivalent to
100% contrast attainable with the equipment
used for L+M). Stimuli rotated at 1 or 10Hz in
randomly
assigned
directions.
Stimulus
generation was accomplished as described in
section 2.1.2 and illustrated in Figure 2.1.
83
EXPERIMENT 3 - Colortuning
Methods
Maximally attainable RMS-cone-contrast
14% 15% 95% 96% 10% 18% 22% 23% 19% 171% 12% 12% 90%
100%
30%
10%
3%
of max
value
Figure 5.2. Horizontal order: Maximally attainable RMS-cone-contrast for 13
color-directions used. Vertical order: Contrast stage values and their perceptual appearance. As
cone-contrast decreases colors desaturate. Note, the cone-contrast stimuli were generated with
respect to the calibration of the projector used in the fMRI experiment. Hence saturation and
detectability might be compromised in this figure. For instance, the lowest L-M stimulus contains 0.3%
cone-contrast, which is proximally double the detection threshold at 1.5Hz temporal frequency and
below detection threshold at 7.5Hz (Liu & Wandell, 2005).
...
Figure 5.3. Exemplification of stimulus time course (presentation of stimulus events),
and the resultant hypothetical BOLD-response time course as could be expected from a voxel located
in a cortical region associated with visual processing. Each run started with 10s fixation task on the
mean gray background only. Stimulus events were presented for 1s. ISIs of mean gray presentation
and fixation task varied between 2s and 6s (mean 2.5s). Presentation of all stimulus colors and
contrast levels were randomized. Each run ended with 55s blank screen presentation.
84
EXPERIMENT 3 - Colortuning
Methods
Figure 5.3 illustrates the stimulus presentation time course and a hypothetical BOLDresponse time course for one voxel, the result of the overlapping BOLD-responses elicited at
each stimulus event. Deconvolution was established using the WMLE-procedure introduced
by Hinrichs et al. and described in section 2.2.3.2 (Hinrichs et al., 2000).
5.2.3 MR-protocol
The subjects were scanned in a GE-Signa LX 1.5T neuro-optimized system with a 5”
surface coil positioned at the occipital pole. BOLD-contrast images were acquired with a
gradient echo planar imaging (EPI) sequence (TR=2s; TE=40ms; flip angle 80deg). 23 slices
were oriented approximately perpendicular to the occipital sulcus (no gap). The voxel size
was 2.81*2.81*3mm (64*64 pixel in-plane matrix). For the anatomical localization of the
functional images T1 weighted images (2D-spin-echo sequence) were scanned with the same
slice position and thickness as the BOLD-images at a higher inplane resolution (0.7*0.7mm,
in-plane resolution 256*256 pixels). For both subjects retinotopic maps were obtained using a
standard protocol (Engel et al., 1994; Engel et al., 1997b) and the software package mrVista
as described in Experiment 1, section 3.2.1 (Teo et al., 1997; Press et al., 2001).
5.2.4 Fitting contrast response curve
In the further analysis only voxels activated by the highest contrast that fell into the
borders of the ROIs under study were included. For each measured scanning session and for
each color-contrast-speed condition, the WMLE-procedure (Hinrichs et al., 2000) yields one
weight for a fixed hemodynamic BOLD-response function which was determined in a data
based procedure, described in section 2.2.3.2. The weights determined this way can be
interpreted as BOLD-amplitudes with the same but arbitrary scaling over conditions. For each
of the thirteen color directions, nine measurement values (weights) were obtained (one for
each session on separate days) for each of the 4 cone-contrast stages and two velocity
conditions. In order to visualize cone-contrast dependence for each color-speed condition we
chose to fit these weights with a sigmoid function for each visual area (Boynton et al., 1996;
Logothetis & Wandell, 2004; Liu & Wandell, 2005). The BOLD-contrast response function
was modelled as follows:
R=M
(c/s)p
(c/s)p + 1
(Eq.: 5.1.),
85
EXPERIMENT 3 - Colortuning
Results
where R is the BOLD response amplitude, c is the presented stimulus contrast, M is the
saturation amplitude (maximal amplitude value), s is the semi-saturation contrast parameter
(mean of all amplitude values) and p is the slope of the sigmoidal function.
Fitting these parameters per color and velocity combination is based on multiple
BOLD-amplitude estimates, which were collapsed over hemispheres for each subject. Every
fit is therefore based on 72 data points (i.e.: 4 cone-contrast levels * 9 days of measurements *
2 hemispheres). Weights for a given visual area from both hemispheres were included into the
fitting procedure giving a total of 18 measurements per color-contrast level and temporal
frequency condition. The matlab function fminsearch was used (Math Works, Natick, MA) to
minimize the sum of squared differences of the sigmoidal function and the obtained BOLDresponses per contrast level. Errorbars for each measured contrast level indicate the standard
deviation derived from 18 underlying measurements. Error tubes have been added to the fitted
curves, which were derived through linear interpolation between each consecutive upper and
lower error range respectively.
5.3 Results
The goal of this experiment was to characterize chromatic properties of human visual
cortex by using an array of 13 cone-contrast colors, spanning all major axes and thus five
planes in cone-contrast space, parameterized in four cone-contrast stages and investigating the
effects of two stimulus velocities. Measurements were conducted on 18 days per subject (9
days*1Hz, 9 days*10Hz). Since only two subjects were available for these kinds of extensive
measurements they were separately analyzed, which in turn allows for a check of the
replicability of the results.
Eight three-way rANOVAs for eight different ROIs (V1, V2d, V2v, V3d, V3v, hV4,
V3A & MT+) parted by subject were conducted. The factors were speed (2 levels, 1 & 10Hz),
color (13 levels) and contrast (4 levels, 100%, 30%, 10% and 3% percent of the maximally
attainable cone-contrast in every color direction). The degrees of freedom were GreenhouseGeisser corrected when appropriate. The Bonferroni-correction for 8 rANOVA-tests yields a
p-value of 0.0063 (0.05/8). In addition, posthoc t-tests will examine specific effects of speed
when called for by the global analysis and specific hypothesis generated through examination
of the contrast response curves of the two speeds for each subjects ROI-color combination.
The corrected p-value for multiple comparisons for follow up t-tests within one ROI is
equivalent to p ~ 0.001 (0.05 / (13 colors * 4 contrasts) = 0.00096154) and over all ROIs p ~
0.0001 (0.05 / (13 colors * 4 contrasts * 8 ROIs) = 0.00012019). For brevity sake only pvalues are reported. For a complete list of all rANOVAs and t-test results as well as the
86
EXPERIMENT 3 - Colortuning
Results
corrected degrees of freedom, please refer to the appendix (Table 5.1 and 5.2). A negative tvalue in the t-test-tables indicates a higher response for 10Hz than 1Hz generated BOLDs.
5.3.1 Global effects of color, contrast and velocity
Without exception, the rANOVAs yielded a highly significant main effect of color in
all ROIs of both subjects (all p < 0.00001), indicating that different colors yield different
responses in all visual areas, which was already reported in Experiment 2 (Figure 4.3).
Similarly, a significant effect is observed for the factor contrast in all ROIs and for both
subjects, which could be expected since BOLD-responses are known to increase as stimulus
contrast is raised (Boynton et al., 1996; Liu & Wandell, 2005).
The effect of speed for each visual area was very similar between subjects. Significant
main effects of speed for both subjects were observed in the following areas: V2d (Dn45: p =
0.002, Lk13: V2d, p < 0.000), V3d (Dn45: p < 0.000, Lk13: p < 0.000), V3v (Dn45: p =
0.005, Lk13: p = 0.001), hV4 (Dn45: p < 0.000, Lk13: p = 0.001) and V3A (Dn45: p < 0.000,
Lk13: p = 0.002. A Bonferroni corrected significant effect in area MT+ was only observed in
subject Dn45 (p = 0.002). This indicates, that speed has a differential effect on the BOLDresponses in some but not all visual areas. For instance, both subjects do not show significant
effects of speed in V1 (p>0.05). The reason becomes clear as one examines the contrast
response functions for the two velocities for a given color direction in V1 (Figure 5.4), which
will be reported in section 5.3.3.1.
5.3.2 Interactions
Bonferroni corrected significant interactions of factors color and contrast were
observed in all ROIs investigated in subject Lk13 (p < 0.000) whereas this interaction was
significant after Bonferroni correction for subject Dn45 only in hV4 (p < 0.000) and V3A (p =
0.002); and uncorrected significant in the following ROIs: V1, p = 0.019; V2v, p = 0.037;
V3v, p = 0.034 and MT+, p = 0.008. This interaction indicates for the visual areas where this
effect was significant that BOLD-responses to different colors are differentially effected by
contrast.
The interaction between velocity and contrast was significant (uncorrected) in all
visual areas for both subject, but after Bonferroni-correction only the following regions
showed a significant effect in both subjects: V1 (Dn45: p = 0.003, Lk13: p < 0.000), V2d
(Dn45: p < 0.000, Lk13: p < 0.000), V2v (Dn45 p = 0.001, Lk13: p = 0.001) and V3d (Dn45:
p < 0.000, Lk13: p = 0.002). The following ROIs yielded additional Bonferroni corrected
87
EXPERIMENT 3 - Colortuning
Results
interactions in subject Dn45: V3v, p<0.000; hV4, p<0.000; V3A, p<0.000; MT+, p<0.000;
but not in subject Lk13.
Finally, the interaction between velocity and color was significant after Bonferroni
correction in very few visual areas under investigation for subject Dn45: hV4, p<0.000; V3A,
p<0.008; subject Lk13: V2d, p < 0.004.
The three-way interaction between velocity, color and contrast was not significant after
Bonferroni correction in subject Dn45 and marginally significant in Lk13 for the following
ROIs: V3d and V3v (p = 0.004) as well as V3A (p = 0.003). The interactions of velocity and
contrast as well as between color and contrast will be explored further in the following section.
5.3.3 Contrast response curves
As noted above, repeated measures ANOVAs resulted in significant interactions
between velocity and contrast as well as interactions between color and contrast in a majority
of visual areas. To explain this effect we now must turn to the contrast response functions
within several visual areas to investigate in what way velocity and contrast or color and
contrast may have a differential effect on the measured BOLD-responses. Due to limited
space and to further facilitate comparisons with experiment 2, result presentation is restricted
to visual area V1, hV4, V3A and MT+.
5.3.3.1 Effects of stimulus contrast and velocity in V1
For all measured visual areas the highly significant main effect of contrast can be
attributed to the monotonic increase in BOLD-response as stimulus contrast increases in all
visual areas (Boynton et al., 1996; Liu & Wandell, 2005). Surprisingly, both subjects’
rANOVAs, yielded no significant main effect of speed in area V1 but an interaction between
speed and contrast. The different panels in Figure 5.4 represent contrast response curves for
two speeds. The cone-contrast directions for displayed data were chosen to be the same as
measured in Experiment 2 in order to facilitate comparison. Velocity differences for all other
measured cone-contrast space directions can be judged in Table 5.2 of the appendix. Steeper
increases for the high velocity contrast response curves for all color-conditions lead to a clear
intersection between the two speed curves in every panel, indicative of an interaction between
speed and contrast in V1. As stimulus contrast increases responses to high-speed stimuli
exceed responses to low speed stimuli in area V1. The contrast range in which this
intersection takes place differs over color directions.
Statistical comparisons between high and low speeds at the highest contrast values
measured, endorse this effect. For instance, when high-speed responses exceed low speed
88
EXPERIMENT 3 - Colortuning
Results
responses in our analysis, t-values will be negative (Table 5.2 appendix). Comparing both
speed responses at the highest cone-contrast value yields negative t-values in both subjects
and for all color directions. Comparing both speed responses at the lowest cone-contrast value
yields positive t-values in both subjects and for all color directions. This is in concordance
with the Bonferroni corrected significant interaction effect observed in V1. A response profile
where high speed paired with high contrast elicits higher BOLD-responses as compared to
low speeds and high contrasts in all color directions was unique to V1.
V1
Figure 5.4: Contrast Response Function for area V1 in Subjects Dn45 and Lk13.
Increasing cone-contrast yields increasing BOLD-responses. The X-axis uses a log scale to represent
stimulus contrast. The Y-axis represents percent BOLD-signal change as measured using the WBA
(see Methods). Black and red circles represent average data for 1Hz and 10Hz respectively, measured
on 9 days for each color/contrast-combination and each speed separately. Black and red lines indicate
1Hz and 10Hz contrast response functions respectively. Errorbars indicate 1 standard deviation. Note
that the maximally reached cone-contrast differs over color directions (x-axis). The blue dotted lines
approximate 10% cone-contrast in all figure panels (the highest cone-contrast that could be reached in
the L-M cone-contrast direction). This contrast value will be used as comparison value for color
selectivity investigations over visual areas in section 5.3.4.
5.3.3.2 Effects of stimulus contrast and velocity in hV4 & V3A
Compared to V1 contrast response curves for area hV4 (Figure 5.5) and V3A (Figure
5.6) exhibit a markedly different response profile while response pattern for hV4 and V3A are
remarkably similar. This provides further support for the results from experiment 2 with
respect to the similarity of the response properties of area V3A and hV4. Results of
experiment 2 additionally suggested a preference for slowly moving chromatic targets in these
regions, while a recent report assigns a high-speed achromatic preference to V3A (Liu &
Wandell, 2005). It is thus of particular interest to compare the response profiles elicited by the
89
EXPERIMENT 3 - Colortuning
Results
two velocities for the L-M and L+M+S stimuli in V3A and hV4, because they emphasize a
chromatic R/G- and achromatic B/W-mechanisms respectively.
Contrarily to the results from Liu et al. (2005) we find no support for a fast speed
preference in these regions, because for subject Dn45, no Bonferroni corrected significant
velocity differences at any cone-contrast stage of achromatic L+M+S stimulation was found
in neither hV4 nor V3A (p>0.001). Furthermore, the V3A response profile to achromatic
stimuli for subject Lk13 shows a Bonferroni corrected significant difference at the lowest
contrast stage of 5% (p<0.001), but in the reverse direction as expected from Liu et al. (2005),
indicating low speed preference, while no other achromatic cone-contrast stage yielded
significant velocity differences in V3A. Similarly in hV4 for subject Lk13, stimuli at the
lowest achromatic cone-contrast stage yielded a significant velocity difference after
Bonferroni correction (p<0.001), while no other contrast stage showed a corrected significant
differences (p>0.001) indicating a slow speed preference as well.
Having shown that V3A and hV4 show no preference for fast achromatic speed, we
may test predictions from experiment 2, which suggested a slow speed chromatic preference
in these regions. A slow speed chromatic preference for both regions becomes obvious as
responses to R/G stimuli (L-M) are compared at low cone-contrast stages (Figure 5.5 and 5.6).
For instance, responses to L-M stimulation in both regions clearly differed between the low
and high-speed condition up until the third contrast stage of 3% for subject Lk13 after
Bonferroni correction (p<0.001). For subject Dn45, the response difference elicited by the two
velocities was Bonferroni corrected significant until the third contrast stage for V3A
(p<0.0001) and for hV4 until the second contrast stage (1%), Bonferroni corrected
(p<0.0000). At 10% cone-contrast, the highest L-M contrast possible, both speeds generate
roughly equal responses in both areas. These response profiles clearly indicate a low speed
chromatic preference for both hV4 and V3A in concordance with experiment 2. In addition, a
response profile as observed in V1 where the high-speed contrast response curves exceed the
low speed curves at sufficiently high contrasts was not prevalent in either hV4 or V3A.
It should be noted that both areas gave strong responses at the two lowest conecontrast stages for the chromatic conditions L-M and pure S-cone stimuli at low velocity,
indicating a very high sensitivity to low chromatic contrast in area hVA and V3A. This differs
to response profiles measured in motion sensitive cortex MT+, which are presented in the next
section.
90
EXPERIMENT 3 - Colortuning
Results
Figure 5.5: Contrast Response Function for hV4 in Subjects Dn45 and Lk13.
Plotting conventions as in Figure 5.4
Figure 5.6: Contrast Response Function for V3A in Subjects Dn45 and Lk13.
Plotting conventions as in Figure 5.4
91
EXPERIMENT 3 - Colortuning
Results
5.3.3.3 Effects of stimulus contrast and velocity in MT+
A yet different contrast response profile was observed for visual area MT+ (Figure
5.7). In both subjects, the two contrast response curves per speed and color direction were
more similar (closer) to one another with respect to response magnitude as compared to the
two velocity contrast response functions in area hV4 and V3A. This effect appears to be due
to a general attenuation of low speed responses, observed across all color directions.
Motion sensitive region MT+ is believed to receive visual information mainly by
means of the magnocellular pathway, while parvo- and konio-cellular pathway contributions
have been reported, but seem not fully established (Maunsell et al., 1990; Gegenfurtner et al.,
1994; Thiele et al., 1999; Wandell et al., 1999; Nassi & Callaway, 2006). It is thus of interest,
to compare response differences elicited by color-directions, which are assumed to emphasize
these pathways (Krauskopf, 2000a) and in addition investigate possible velocity effects.
Color-directions, which are believed to be primarily transmitted through the
magnocellular pathway, are L+M and L+M+S (Gegenfurtner & Sharpe, 1999). Testing
velocity differences between the two speed conditions at each cone-contrast stage for
perceptually achromatic L+M+S stimuli with t-tests, reveals no significant difference after
Bonferroni correction for both subjects (p>0.01). Indicating a speed invariant response profile
to achromatic stimuli in area MT+. Testing speed differences for stimulation modulated along
the L+M cone-contrast direction in both subjects revealed also no significant velocity
difference after Bonferroni correction (p>0.01), for all but the lowest cone-contrast stage in
subject Lk13 (p<0.001). The contrast response curves in Figure 5.7 illustrate these effects.
Chromatic stimulation, typically associated with parvo-cellularly driven transmission,
is best modulated along the L-M-axis in cone-contrast space (Derrington et al., 1984).
Comparing responses at each cone-contrast stage over velocities, reveals no Bonferroni
corrected significant differences for subject Lk13 (p>0.01), while at the lowest L-M-conecontrast stage comparison (0.3%), a significant difference was observed for subject Dn45
(p<0.0001). It is worth noting that a strong slow-speed response produced this difference.
Stimulation emphasizing the konio-cellularly driven path, while keeping L+M
stimulation constant, modulates the S-cones only, representing yet another purely chromatic
cone-contrast direction. Testing velocity response differences along the S-cone axis in conecontrast space reveals no Bonferroni corrected significant difference (p>0.001), for both
subject at all but the lowest cone-contrast stage of 2.85% (p<0.0001) in subject Dn45. Again,
this effect was produced by a slow-speed response.
In summary, all but a few cone-contrast directions and those inconsistent across
subjects revealed no significant effects of velocity compared at the respective cone-contrast
92
EXPERIMENT 3 - Colortuning
Results
stage in area MT+. If comparisons were significant at a Bonferroni corrected level, they
tended to be for very low cone-contrast values. The relative speed invariance as reported in
experiment 2 can thus be confirmed.
Figure 5.7: Contrast Response Function for MT+ in Subjects Dn45 and Lk13.
Plotting conventions as in Figure 5.4
5.3.4 Comparison at equal cone-contrast
The analysis of BOLD-responses in experiment 3 has so far focused on cone-contrast
directions, which emphasize the stimulation of known post-receptoral mechanisms while
keeping contribution from other mechanisms constant. We chose to compare the responses at
several cone-contrast levels for these color directions in experiment 3, to uncover sensitivity
differences in several visual areas, expected from psychophysical studies, and to confirm
predictions made in experiment 2, regarding the association of several visual areas to
postulated mechanisms of motion perception. It is however of high interest to study the
BOLD-responses to cone-contrast directions that do not emphasize only one single
mechanism but lie in between cardinal directions, and thus, stimulate several mechanisms
simultaneously 3 . This is of high interest because these cone-contrast combinations span
planes in cone contrast space, which have previously not been investigated.
We measured thirteen cone-contrast space directions at individually scaled conecontrast levels, the responses of which have been interpolated with a sigmoid interpolation
3
For simplicity, these cone-contrast directions are referred to as ‘mixed colors’
93
EXPERIMENT 3 - Colortuning
Results
function for each color and velocity combination separately (see section 5.2.4). We refrain
from presenting cone-contrast response functions to all of these directions due to limited
space, and chose to present responses to all cone-contrast direction at an equal cone contrast
for visual areas V1, hV4, V3A and MT+ (see sections 1.2.3.6, 1.3 and 2.1.2). This
characterizes chromatic selectivity within one visual area for equal physical contrast at the
initial input stage (see section 1.2.1.1) and provides clues, how chromatic selectivity may
change as we compare effects of velocity or color within and across areas.
Statistical analysis is omitted at this point, because the responses shown are
interpolated values determined at the 10% cone contrast point shown as blue dotted lines in
figures 5.4 to 5.7. Error bars can be regarded as a measure for reproducibility since 18
measurements went into every value (9 scanning sessions & 2 hemispheres, see Section
5.2.4). The 10% cone contrast value was chosen as the color sensitivity comparison value,
because it was the highest achievable equal cone contrast in the L-M direction, which in
addition lay well within the range of all other achievable colors of the equipment and assured
measurable responses in all visual areas at both speeds. Comparable color conditions, actually
measured at cone-contrast stages fairly close to the 10% criterion, are as follows: L+M
(9.6%), S (9.5%), S-L-M (9%), L-M (10%), L-M+S (12%), L-M-S (12%), L (14%), M (14%)
and L+M+S (17%).
5.3.4.1 Effects of stimulus velocity and color
This section compares BOLD-responses at equal 10% cone-contrast, similar to
experiment 2, in order to assess chromatic selectivity and the effect of speed in the visual
areas. We find that for all ROIs and for both subjects the effect of color is more pronounced
than the effect of speed, which means that the effect of speed does not override the effect of
color. This becomes clear as one examines Figures 5.8 to 5.11.
At 10% cone-contrast, one can for instance apply the following general rule: If a 1Hz
response in color direction X is stronger than the response to 1Hz stimuli in color direction Y,
the same response relation will be observed for 10Hz stimuli. The values for factor speed are
exchangeable in this rule. In other words color leads speed with respect to magnitude of
BOLD-responses.
Color selectivity profiles measured for several visual areas in both subjects are highly
similar for the same areas in both subjects. The effect of stimulus color in V1 is quite striking.
Figure 5.8 reveals that at a cone-contrast of 10%, V1 responses to stimuli, putting emphasis
on the B/Y mechanisms (S) and achromatic mechanism (L+M+S) are smallest for both
speeds, while R/G modulations (L-M) exceed them by far. The response to luminance stimuli
94
EXPERIMENT 3 - Colortuning
Results
(L+M) exceed responses to stimuli emphasizing the achromatic mechanism (L+M+S) in both
subjects, indicating a probable inhibitory property of S-cone contribution on the BOLDresponse, which is also indicated by responses to stimuli containing pure S-cone contribution,
opposed by the luminance channel (L+M). Responses to those colors (S-(L+M)) are also
smaller than to L-M especially at high speeds (reproduced as in Experiment 2).
In addition, the expected response magnitude hierarchy (L-M > L+M > S-responses)
can be confirmed at both speeds for V1 neuronal populations (Engel et al., 1997b; Liu &
Wandell, 2005). The higher responses to fast moving colors, reported in section 5.3.3.1, can
already be observed for some color directions, while some need a higher contrast stage to
show this effect (Figure 5.8).
Figure 5.8: V1- BOLD-responses at 10% equal cone-contrast
Data are shown for all 13 cone-contrast-space directions for 1Hz (blue bars) and 10Hz (red bars)
speeds. Subject Dn45 upper panel ; subject Lk13 lower panel. Error bars indicate the standard error
derived from 18 values (see section 5.2.4) at 10% cone-contrast determined from the each separately
fitted contrast response function. The lowest responses were obtained for stimuli that emphasize
putative mechanisms except for L-M (R/G-mechanism). Responses to high velocities exceed
responses to low velocities for some color directions. As contrast is raised all responses to high
velocities will exceed the low speed response (see contrast response curves in section 5.3.3.1).
95
EXPERIMENT 3 - Colortuning
Results
Figure 5.9: hV4 BOLD-responses at 10% equal cone-contrast
Conventions are the same as in Figure 5.8. Compared to 10% cone-contrast responses in V1, hV4
responses at slow speeds are substantially amplified, while high-speed responses are always lower
than low speed responses.
Figure 5.10: V3A BOLD-responses at 10% equal cone-contrast
Conventions are the same as in Figure 5.8. Compared to 10% cone-contrast responses in V1, V3A
responses at slow speeds are substantially amplified, while high-speed responses are always lower
than low speed responses.
96
EXPERIMENT 3 - Colortuning
Results
Comparing equal 10% cone-contrast responses in areas hV4 and V3A in Figures 5.9
and 5.10 to V1 (Figure 5.8), the picture changes substantially, while response profiles for hV4
and V3A are highly similar. The above mentioned sensitivity hierarchy of response magnitude
in the order of L-M > L+M > S, is well pronounced at 10Hz stimuli but weaker for 1Hz
stimuli. Responses to 1Hz moving patterns appear more similar in both regions. This
emphasizes hV4 and V3A as highly sensitive to all color combinations and clearly indicates
their preference for low speeds because responses to slow moving colors are amplified
compared to V1, while fast moving color responses mirror the V1 profile.
Figure 5.11: MT+- BOLD-responses at 10% equal cone-contrast
Conventions are the same as in Figure 5.8. Compared to 10% cone-contrast responses in hV4 and
V3A, MT+ responses at slow and high speeds are roughly equal.
The color and speed response profile is different to hV4 and V3A in a way that slow
speed preferences are far less prevalent especially for achromatic stimuli (Figure 5.11). While
in hV4 and V3A fast achromatic responses are well below slow chromatic responses, this
difference is marginal in area MT+. We thus reproduced results from experiment 2, in which
we found no achromatic fast speed preference in these regions but a slow chromatic
preference. In addition, a preference in MT+ for stimuli emphasizing a luminance or
achromatic mechanism, as expected from a region, assumed to be primarily driven by
magnocellular signals (Maunsell et al., 1990) cannot be confirmed, as responses to L+M,
L+M+S and L-M stimuli are compared. Responses to these colors are found to be roughly
97
EXPERIMENT 3 - Colortuning
Results
equal to one another and to both speeds, indicating a relative indifference to cone-contrast
direction. However, the well described weak response to pure S-cone stimuli in MT+, can be
confirmed here as well (Seidemann et al., 1999; Wandell et al., 1999), especially at high
temporal frequencies.
Interestingly, for both subjects and in all visual areas measured, highest responses are
evoked by ‘mixed colors’, which may put emphasis on several cortico-geniculate channels
simultaneously. The effect is most pronounced for color directions emphasizing the B/Y- and
R/G- mechanism simultaneously (L-M+S & L-M-S). If two different color mechanisms are
stimulated simultaneously, a fairly high BOLD-signal is expected, since separate cell
populations tuned to those mechanisms in the respective visual areas, will respond (but see
discussion).
5.3.5 Ellipse fits
5.3.5.1 Effects of stimulus contrast, color and velocity
In order to derive a more complete understanding of human cortical color processing
and its dependence on stimulus speed, this section aims to summarize the data for all color
directions in the contrast range. Due to limited space, a presentation of cone-contrast response
curves to all used color directions was omitted in section 5.3.3, and the presented bar plots in
section 5.3.4 do not allow to judge the effects of increasing contrast over all color directions.
To exemplify the relationship between cone-contrast and all measured color-directions, ellipse
fit plots have been constructed, in which over 4 equal-cone contrast stages, the effects of
stimulus contrast can be comparably judged across all color directions within one ROI in
addition to the effects of velocity (Figures 5.12 to 5.15). Similar plots have been provided by
Engel et. al. (1997), who sampled responses in two cone-contrast planes (Engel et al., 1997b).
Responses to stimuli in the same two planes used by Engel et.al (1997) appear in the two left
most columns, while the additional three planes from this study appear in the three right most
columns. Since measurements in primary visual cortex revealed the most unique effects of
velocity over contrasts for all color directions, we chose to present V1 ellipse fit plots for both
subjects. To underline the uniqueness of V1 contrast and speed effects we provide the same
ellipse-fit plots for the motion sensitive region MT+.
Each row in Figures 5.12 to 5.15 represents BOLD-responses at increasing equal conecontrast of 2.5%, 5%, 7.5% and 10% respectively. The higher contrast levels are not displayed
because no data could be obtained for contrasts higher than 10% in the L-M direction. All
values are based on fitting equation 5.1 to the responses at every cone-contrast level for every
98
EXPERIMENT 3 - Colortuning
Results
scanning session and for each ROI/color/contrast condition, which were intercepted at the
requested contrast values (2.5%, 5%, 7.5% and 10%). The average of eighteen values
obtained this way is plotted along the appropriate axis, representing measurements from two
hemispheres at nine days for 1Hz and 10Hz respectively.
S
L-M
Constant plane
M
L+M
L
2.5%
Equal cone contrast
5%
7.5%
10%
Figure 5.12: Ellipse-fit-plots of V1 BOLD responses, subject Lk13.
Each panel represents BOLD responses in one plane of cone-contrast space at equal cone-contrast.
Each row represents equal cone-contrasts in five planes: Rows: 1st: 2.5%-; 2nd: 5%-; 3rd:7.5%-; 4th:
10% equal cone-contrast. Each column represents the same cone-contrast plane in which
contributions from specific cone-classes or combinations of them are kept constant: Columns: 1st:
constant S-; 2nd :constant L-M-; 3rd : constant M-; 4th : constant L+M-; 5th : constant L-plane. BOLDresponses within each plane are displayed along four cone-contrast space directions depicted by gray
arrows. Distance from the coordinate systems center represents BOLD-response magnitude plotted on
the appropriate color axis. The circles represent interpolation values, black and red for 1 and 10Hz
respectively. BOLD-response values are fitted with an ellipse: Black lines 1 Hz; Red lines 10 Hz
responses. BOLD-responses increase with increasing cone-contrast along rows. Responses to fast
moving stimuli exceed responses to slow moving stimuli, but the contrast values at which this occurs
differs across cone-contrast space directions. Highest responses are observed along directions
opposing L-M and S-cone signals. The ellipse area in the constant L-M planes (2nd column) is
substantially smaller as compared to MT+ (Figures 5.14 and 5.15). For further details see section
5.3.5.1.
Each column represents stimuli, from four color direction, that lie within one plane of conecontrast space. In the first column, for example, x- and y-axis represent pure L-cone and pure
M-cone stimuli respectively. The luminance mechanism is along the L+M diagonal because it
is assumed to sum L- and M-cone signals. The R/G-mechanism is represented by the L-M
diagonal, because it takes the difference between L- and M-cones, representing an opponent
color mechanism. S-cone stimulation is held constant within this plane and would form the z99
EXPERIMENT 3 - Colortuning
Results
axis, pointing towards the reader. The plane in the second column keeps L-M contributions
constant, the third column for instance keeps M-cone contributions constant, the forth column
keeps the luminance contributions constant (L+M), while within the fifth column, L-cone
influence is constant.
As contrast increases from the first to the fourth row BOLD-responses increase. These plots
are well suited to judge, for example the contrast value at which high- speed stimuli exceed
slow speed stimuli in the different color directions.
S
L-M
Constant plane
M
L+M
L
2.5%
Equal cone contrast
5%
7.5%
10%
Figure 5.13: Ellipse-fit-plots of V1 BOLD responses, subject Dn45.
Plotting conventions as in Figure 5.12. BOLD-responses increase with increasing cone-contrast along
rows. Responses to fast moving stimuli exceed responses to low moving stimuli, but the contrast
values at which this occurs differs across cone-contrast space directions. Highest responses are
observed along directions opposing L-M and S-cone signals at 10% equal cone-contrast. For further
details see section 5.3.5.1.
For instance in Figure 5.12 (V1) it can be seen that while the BOLD-response elicited
by fast L+M stimuli is below the response to the slow stimuli at 10% cone-contrast (4th row),
a fast pure L-cone stimulus elicits similar responses to slowly moving L-cone patterns at 5%
cone-contrast but exceeds them at 7.5% cone-contrast. The same response profile where
responses to fast moving stimuli exceed responses to slow moving stimuli is observed for the
L-M (R/G) mechanism. Interestingly for both subjects V1s, ellipse fits in the second column
reveal that the area spanned by the fitted ellipses (i.e. BOLD-responses) is relatively small
when the contrast in L-M direction is set to zero, suggesting a stronger contrast gain for the LM mechanisms as compared to the pure S-mechanism.
100
EXPERIMENT 3 - Colortuning
Results
The pattern of results changes substantially in MT+ for both subjects (Figure 5.14 and
5.15). In MT+, the BOLD-response to additive cone signals such as L+M and L+M+S
increase more strongly with increasing contrast as compared to V1. In both subjects, L+M and
L+M+S response magnitude in MT+, exceed V1 responses to these stimuli at the lowest
contrast level. This can be interpreted as indicator for higher contrast sensitivity to luminance
signals in MT+ as compared to V1.
S
L-M
Constant plane
M
L+M
L
2.5%
Equal cone contrast
5%
7.5%
10%
Figure 5.14: Ellipse-fit-plots of MT+ BOLD responses, subject Lk13.
Plotting conventions as in Figure 5.12. BOLD-responses increase with increasing cone-contrast along
rows. Responses to fast moving stimuli do not exceed responses to low moving stimuli. Highest
responses are observed along directions opposing L-M and S-cone signals at 10% equal conecontrast. The ellipse area in the constant L-M planes (2nd column) is substantially larger as compared
to V1 (Figures 5.12 and 5.13) For further details see section 5.3.5.1.
As we compare V1 and MT+ with respect to cone-contrast-planes not investigated in
prior human imaging reports (three right columns in Figures 5.12 – 5.15), higher response
magnitudes are observed in MT+ for both speeds at every cone-contrast level, indicating a
motion signal amplification as it is transmitted from primary visual cortex to MT+. It appears
that no response to fast moving stimuli exceeded slow moving stimuli in MT+ as was
observed in V1. This holds for all cone-contrast directions tested in MT+.
The most interesting similarity between V1 and MT+ for both subjects is that
considerable contrast-dependent changes occur in responses to directions with L- and M-cone
contributions of opposite sign underlining the pivotal role of the L-M mechanism in early
visual cortex as well as higher tier areas. Indeed the most extensive increases with contrast
101
EXPERIMENT 3 - Colortuning
Results
were seen in the plane containing the L-M stimuli opposed by the pure S-cone stimulation (4th
column), constituting essentially a stimulus emphasizing the L-M- and S-mechanisms while
leaving the luminance mechanism constant. These stimuli elicit highest BOLD-responses at
very low contrasts in V1 as well as in MT+ of both subjects. In addition, neural populations
underlying these high BOLD-signals are highly motion-sensitive, indicated by high responses
for high and low speeds starting as low as 2.5% contrast. From data reported here it can
however not securely be concluded whether responses to L-M opposed by S-cone stimuli are
generated by a unique cell population tuned to these kind of stimulus configurations, or if
indeed the high BOLD-responses are a consequence of several cell populations responding
simultaneously, such as L-M and S responsive cells.
S
L-M
Constant plane
M
L+M
L
2.5%
Equal cone contrast
5%
7.5%
10%
Figure 5.15: Ellipse-fit-plots of MT+ BOLD responses, subject Dn45.
Plotting conventions as in Figure 5.12. BOLD-responses increase with increasing cone-contrast along
rows. Responses to fast moving stimuli do not exceed responses to low moving stimuli. Highest
responses are observed along directions opposing L-M and S-cone signals at 10% equal conecontrast. The ellipse area in the constant L-M planes (2nd column) is substantially larger as compared
to V1 (Figures 5.12 and 5.13) For further details see section 5.3.5.1.
102
EXPERIMENT 3 - Colortuning
Discussion
5.4 Discussion
The third experiment investigated BOLD-response differences in several visual areas
as a function of speed, cone-contrast and color. The results of which point towards
preferences in separate visual areas for specific stimulus combinations of velocity,
chromaticity and contrast while no visual area is solely designated for either single stimulus
attribute. It was found that hV4 and V3A respond strikingly similar but clearly distinguishable
to contrast response profiles observed in area MT+.
While MT+ responses were roughly equal at high and low velocities, especially for
additive cone signals over a wide range of contrasts, hV4 and V3A respond substantially
stronger to slowly moving low chromatic contrast patterns compared to fast moving contrast
patterns at low contrasts. Furthermore, since slow speed responses at low chromatic contrasts
in V3A and hV4 exceeded those of MT+ for the same conditions, an additional criterion to
associate visual areas to psychophysically derived motion perception channels, stating higher
chromatic sensitivity in cortical regions associated with the slow motion channel, is fulfilled,
and the conclusions made in experiment 2 can be consolidated. Areas V3A and hV4 are very
likely to account for the slow motion channel of motion perception, while MT+ is likely to
account for the fast motion mechanism.
Primary visual cortex was found to exhibit yet another contrast response profile.
Responses to fast moving stimuli exceed those of slowly moving stimulus patterns at
sufficiently high cone-contrast levels, not observed in any other region. The cone-contrast
value at which higher speeds exceed responses to lower speeds depends on the cone-contrast
direction of the stimulus.
5.4.1 Neurophysiological studies and speed in V1
As suggested in Experiment 2 for V1, stimuli comprised of certain color and contrast
combinations can yield higher BOLD-responses to fast speeds as compared to slow speeds. In
concordance to that view, experiment 3 suggests that the ratio of slow and high speed BOLDresponses is not constant over different contrast levels and may even reverse with increasing
contrast. The contrast range, at which this reversal speed preference takes place, differs
between color directions, as we compare BOLD-responses to low (1Hz) and high-speed
(10Hz) stimulation in experiment 3. Note that in experiment 2, the speed response profile to
L-M stimuli already suggested such an effect. In experiment 2 where three speed conditions
were investigated, it was found for L-M stimuli, that responses increased as speed increased
while cone-contrast was constant at 14% cone-contrast. Experiment 3 extends this finding by
103
EXPERIMENT 3 - Colortuning
Discussion
showing that specifically in V1, high-speed responses can exceed slow speed responses for all
color directions provided that a sufficiently high cone-contrast is applied.
For the L-M mechanism, a 14% cone-contrast employed in Experiment 2, is more than
sufficient to elicit higher responses to fast speeds as compared to slow speeds, because as can
be seen in Figure 5.4, interception occurs already at around 10% cone-contrast for this color
direction. None of the other colors in Figure 5.4, which are the same as in experiment 2, show
this kind of behavior until a cone-contrast of roughly 20% is reached while pure S-cone
stimuli must reach nearly 30% cone-contrast to yield higher responses to high speeds
compared to low speeds. From the results of experiment 3 we would expect no other color
stimulus besides L-M in Experiment 2 to show this kind of response pattern, but since all
cone-contrast space directions clearly showed an intersection of the two velocities contrast
response curves at some cone-contrast value, a general property of V1 may have been
uncovered. This finding underscores that it is highly important to take the contrast level into
account when conclusions about speed preferences are drawn, at least for human V1. This is
not only important when responses are compared over different directions in cone contrast
space but even when stimuli with the same direction in cone contrast space are used. The
conclusions may depend heavily on the contrast level employed. But what might be the
neurophysiological underpinnings of this phenomenon?
V1 receives its bottom up input from the LGN, making it possible that the specific
velocity tuning changes over contrast levels towards higher frequencies may reflect the LGN
input signal. It has however been shown in cats and monkeys, that LGN neurons are able to
follow much higher temporal frequencies than V1 Neurons (Movshon et al., 1978; Orban et
al., 1985; Hawken et al., 1996), and extensive low-pass temporal filtering between primary
visual cortex and LGN has been proposed (Movshon et al., 1978; Orban et al., 1985; Hawken
et al., 1996). These findings suggest that it is likely that a contrast dependent speed tuning
evolves within V1 neuronal connections. Further support for the existence of a contrast
dependent speed tuning effect in V1 neuronal networks, comes from studies, investigating
temporal frequency tuning of cortical neurons as a function of stimulus contrast in non-human
primates (Holub & Morton-Gibson, 1981; Foster et al., 1985; Albrecht, 1995; Hawken et al.,
1996; Kayser et al., 2001; Alitto & Usrey, 2004). A direct comparison of the conclusions of
two of these studies, investigating temporal frequency tuning in monkey V1 neurons by
applying substantially different contrast values, underscores the influence stimulus contrast
can have to measurements of temporal frequency tuning in monkeys. An early study
investigating cortical temporal frequency selectivity in macaque V1 and V2 (Foster et al.,
1985), using a contrast level of around 35% found the preferred temporal frequency tuning in
104
EXPERIMENT 3 - Colortuning
Discussion
V1 cells to be in the range of 5.6-8Hz. A more recent study, applying a much higher contrast
level of 65%, measuring temporal frequency tuning in macaque primary visual cortex and
LGN, find higher preferred peak temporal frequencies for simple-, complex- and LGN-cells to
be 9.5, 11.2 and 16Hz respectively (Hawken et al., 1996). The direction of the speedtuning
distribution shift is rightward, indicating an increase of preferred temporal frequencies in V1
neurons with increasing stimulus contrast, which is in concordance with the data presented in
experiment 3. We showed that higher stimulus contrast leads to higher responses to fast
stimulus speeds as compared to slow stimulus speeds in primary visual cortex of humans
(Figure 5.4).
A reported contrast gain control (CGC), as for instance originally used to describe
retinal ganglion cells of cats (Shapley & Victor, 1978, 1981), could be characterized as
contrast-dependent shift in temporal frequency tuning (Benardete et al., 1992 ; Yeh et al.,
1995; Kremers et al., 1997; Benardete & Kaplan, 1999; Usrey & Reid, 2000). Besides the
known CGC-mechanisms in LGN and retina (Yeh et al., 1995a; Kremers et al., 1997), recent
reports suggest additional cortical nonlinearities to be able to explain the contrast dependent
increase in cortical responses to high temporal frequency stimuli (Carandini et al., 1997;
Kayser et al., 2001; Alitto & Usrey, 2004). A simple increase in contrast of fast moving
gratings causes a rightward shift in temporal frequency tuning of simple and complex cells in
primary visual cortex of cats (Holub & Morton-Gibson, 1981; Albrecht, 1995; Kayser et al.,
2001), ferrets (Alitto & Usrey, 2004) and monkeys (Albrecht, 1995), while it has never been
demonstrated in human visual cortex.
Contrast response curves presented in this thesis lead us to speculate, that human
primary visual cortex is affected by a contrast gain control mechanism, in which high contrast
stimuli moving with reasonably high velocity may be cortically enhanced. Data presented in
Figure 5.4 shows, that the contrast values, at which the contrast response curves for high and
low speeds intersect, differ between color directions. As stated above, the intersectioncontrast-value ranges from ~10% (L-M) to at least ~30% cone-contrast (S). Since the applied
color conditions emphasized specific geniculo-cortical channels it is possible, that this
proposed CGC-mechanism may act differently on these channels, which could be
demonstrated in anesthetized monkeys (Solomon & Lennie, 2005). In addition, it is also
feasible, that distinct layers in primary visual cortex, containing separate cortical termination
sites for the geniculo-cortical channels (Chatterjee & Callaway, 2003), exhibit differential
contrast gain properties which we may have been able to measure, because our stimulation
was designed to put emphasis on these channels. The argument of differential contrast gain
properties depending cortical layer, is supported by the finding of Hawken and colleagues,
105
EXPERIMENT 3 - Colortuning
Discussion
who report the integration time for simple cells in V1 to be highly variable at different levels
of cortical processing (layers) (Hawken et al., 1996). They conclude that V1 cells may not
simply inherit their tuning from LGN and emphasize the existence of a cortical contrast gain
control mechanisms (Solomon & Lennie, 2005), which, as the authors stress, may be
polysynaptic and reflects temporal integration through a cortical network (Hawken et al.,
1996).
The former argument goes well in line with the neurophysiological finding, that high
stimulus contrast speeds up the responses in V1 cells, causing high contrast responses to occur
earlier in the stimulus generated processing cycle (Dean & Tolhurst, 1986; Albrecht, 1995).
We speculated that this could consequently allow the effective cortical processing within the
proposed polysynaptic cortical network (Hawken et al., 1996) to occur at faster and higher
rates when stimuli are presented at high contrasts and fast speeds. If an intracortical network
with elaborate intrinsic connections, like V1, is highly active because stimulus properties such
as high temporal frequency and high stimulus contrast demand faster processing at a given
stimulus presentation time, a higher metabolism is consequently reflected in higher BOLDsignals as compared to stimulus combinations of low speed and low contrast, which is what
we observe.
5.4.2 Neurophysiological studies and color in V1
The most heterogeneous response profile regarding speed and color responses at equal
cone-contrast was observed in area V1 (Figure 5.8). To be specific, response magnitude in V1
elicited by 10% cone-contrast stimuli that put emphasis on a single cortico-geniculate channel,
such as L+M+S or S, are found to be substantially lower as cone-contrast colors that may
emphasize several of these channels simultaneously (Derrington et al., 1984; Lennie, 1988;
Lennie et al., 1990), which, for simplicity reasons, we chose to refer to as ‘mixed colors’, if
not otherwise indicated. Response magnitude to stimulation emphasizing the R/G-channel (LM), however, was found to be equally high as compared to ‘mixed colors’. It appears as if
cortical representation of R/G-, B/Y- and achromatic stimuli may fundamentally differ.
If the assumption is employed that oxygen consumption and related blood volume
changes, the basis for the presented functional imaging data (Ogawa et al., 1990; Ogawa et
al., 1992), is equal across all active V1 cells, a most parsimonious explanation for varying
response magnitudes in V1 may be generated. This explanation would state that primary
visual cortex contains separate neuronal populations, which are tuned to the stimulus
configurations of the geniculo-cortical channels but differ in total cell count.
Neurophysiological investigations on non-human primates have indeed found that V1
106
EXPERIMENT 3 - Colortuning
Discussion
contains separate populations of achromatic, R/G and B/Y color-opponent neurons
(Livingstone & Hubel, 1984; Thorell et al., 1984; Ts'o & Gilbert, 1988; Landisman & Ts'O,
2002), while the approximate number comprising separate populations was different
(Landisman & Ts'O, 2002; Chatterjee & Callaway, 2003).
For instance, Chatterjee et al. encounter nearly double as many achromatic versus R/G
cortical projection sites in macaque primary visual cortex. If the above assumption of equal
metabolism across different cell populations holds, this observation leads us to expect a
BOLD-response magnitude for achromatic stimuli (L+M+S) that is roughly twice as large as
compared to R/G-stimuli (L-M) in human visual cortex. We and other imaging studies find
the contrary, considering L+M stimuli as achromatic for now (Kleinschmidt et al., 1996;
Engel et al., 1997b; Engel & Furmanski, 2001; Liu & Wandell, 2005). The typical BOLDresponse amplitude hierarchy that is measured in human fMRI, is consistently found to be
highest for L-M which is higher than L+M(+S) which in turn is higher than pure S-conestimuli responses at equal cone-contrast levels, which contradicts the reported cell counts.
The above inconsistency between cell count and BOLD-magnitude leads us to
speculate that distinct cell populations in primary visual cortex might substantially differ with
respect to their metabolism. This argument is supported by differential physiology of cells in
different layers of V1 (Livingstone & Hubel, 1988), which may cause possible metabolic
differences between neuronal subpopulations. Specifically the cytochrome-oxidase blobs in
layers 2 and 3, are known to contain specialized vasculature (Zheng et al., 1991), which could
cause neurons therein to be over-represented in BOLD-responses. However, the cell-count
across blobs may be informative and predict measured fMRI responses.
Landisman and Ts'O (2002) found patches in V1 that are preferentially activated by
chromatic gratings compared with luminance gratings. These imaged color patches were
correlated, with the cytochrome-oxidase (CO) blobs, while luminance sensitive cells were
located in between blobs. Interestingly, the results indicate that individual blobs tend to
contain neurons of only one color-opponent cell type: either red/green (L-M) or blue/yellow
(S-(L+M)). It has been reported, that there are more R/G than B/Y color-opponent cells
associated with CO-rich blobs in V1 (Livingstone & Hubel, 1984). If we additionally assume,
that the applied R/G and B/Y stimulation only activates color-opponent cells in the blobs, a
response comparison at equal cone-contrast should yield higher responses to L-M than to S(L+M) stimuli, which we observed for the two subjects contrast response curves of
experiment 3, also confirming results reported in experiment 2.
Nonetheless, to date it is not known in what way different cell populations in distinct
laminar organization of primary visual cortex may differ regarding their energy consumption
107
EXPERIMENT 3 - Colortuning
Discussion
when excited. For instance, BOLD-signals generated by small amounts of cells with a high
metabolic rate could exceed signals from a larger number of cells with low metabolic rate.
The finding that responses to achromatic stimuli are well below L-M color opponent
responses, while their designated cell count may lead us to expect an opposite response
profile, suggests that different neuronal classes in distinct cortical layers differ regarding their
energy consumption. Future neurophysiological work is needed, examining metabolic
properties of different classes of neurons in the laminar structure of V1. This knowledge will
provide a better link between activity of cell populations and the associated BOLD-signal.
Yet another explanation for the observed response differences especially for stimuli
emphasizing cortico-geniculate pathways could be, that V1 may plainly mirror the input level
from LGN-projections because, BOLD-signals are believed to mainly reflect the specific
inputs into a cortical area rather than its spiking output (Logothetis et al., 2001; Logothetis,
2003). The cell ratio for the magno-, parvo-, and koniocellular layers in LGN can roughly be
estimated to be 10%, 80% and 10% respectively, which explains the observed response
profile in human primary visual cortex only poorly. In addition, cells that respond strongest to
stimulation opposing the R/G information with S-cones have not been found in the LGN,
while we measured highest responses to stimuli such as L-M-S and L-M+S (Figure 5.8). We
must thus assume that these kinds of ‘mixed color’ responses in humans emerge in cell
populations of V1.
Clues supporting this claim bear on more recent neurophysiological investigations of
chromatic responses in single cortical cells of the macaque. Conway and Livingstone
confirmed that L-M+S double-opponent cells exist, and Solomon and Lennie found the most
common cortical organization to be L-M-S (Solomon & Lennie, 2005; Conway &
Livingstone, 2006). Given that distinct cell populations can be found in V1, which respond
preferentially to ‘mixed colors’ and we measure BOLD-responses to these stimuli that were
among the highest of all applied colors, these cells must either be very numerous, have an
extraordinary metabolism or another explanation is in order.
Assuming that purely L-M, S, and variations of L-M+S responding cells exist, it is
most likely that the observed high BOLD-responses are the combined sum of all of these cell
populations as all three are presumably excited by the mixed color stimuli described above.
Similarly L-M stimuli could excite L-M-S or L-M+S neurons, which may have added to the
large measured BOLD-responses observed for L-M.
We speculate that a merging of signals from R/G and B/Y responsive cells occurs
somewhere in V1 (Fitzpatrick et al., 1985; Callaway, 2004, 2005; Yoshimura & Callaway,
2005; Briggs & Callaway, 2005 ), possibly by means of the L-M+S responsive populations
108
EXPERIMENT 3 - Colortuning
Discussion
and that both of these stages contribute to the measured BOLD-signal. Evidence from other
studies also supports coupling of M- and S- cones opposed by L-cone input in V1 (Conway,
2001; Wachtler et al., 2003), which was not part of the color stimuli we used.
One may consider using the ‘mixed color’ configurations in future retinotopic
mapping experiments since at 10% contrast BOLD-activations are so much higher than
conventional black and white (L+M+S) responses. This is however not suitable (as of yet),
because additive L-, M- and S-cone stimuli are much easier to generate and even more so, can
reach the highest overall contrast (170% RMS), which in turn will elicit highest BOLDresponses. Nonetheless, as experimental presentation techniques progress it might be possible
in the future to generate L-M-S and L-M+S stimuli that exceed 12% cone-contrast (see visual
stimulation Experiment 2). In order to do that, displays could be designed containing light
sources that enable researchers to maximally avoid simultaneous excitation of the L- and Mcones, in order to increase the L-M cone-contrast. Once achieved, colors could be computed
which will yield higher BOLD-signals as L+M+S at highest contrast levels, because the
‘mixed’ color conditions at 12% cone-contrast (not shown) are already yielding responses that
approach responses from 170% L+M+S cone-contrast, at least in extra-striate regions, which
will be discussed next.
5.4.3 Neurophysiological studies and extra-striate cortex
Response properties of human extra-striate visual areas MT+, hV4 and V3A can be
regarded relatively unexplored territory with respect to cone-contrast controlled stimuli and
the effect of stimulus velocity using fMRI (Wandell et al., 1999; Liu & Wandell, 2005).
Neurophysiological investigations on the macaque monkey have served as a framework to
compare differences in sensitivity to color and motion stimuli of visual areas such as MT
(Maunsell et al., 1990; Gegenfurtner et al., 1994; Seidemann et al., 1999) or V4 (Schein &
Desimone, 1990; Heywood et al., 1992; Yoshioka et al., 1996). To our knowledge, no singleunit study on monkeys exist, investigating cone-contrast responses as a function of temporal
frequency in dorsal area V3A, while some work has been published regarding color
responsivity in macaque V3 (Felleman & Van Essen, 1987; Gegenfurtner et al., 1997) and V2
(Gegenfurtner et al., 1996). This study aims to close that gap and to further advance the
understanding of cortical color processing in humans.
Sensitivity difference of the extra-striate visual areas regarding color or motion
responses have been used to assign these regions to presumed cortical ventral and dorsal
pathways (Zeki, 1974; Ungerleider & Mishkin, 1982; Mishkin et al., 1983; Zeki, 1983c).
While color was assumed to be the critical stimulus property of the ventral WHAT-path,
109
EXPERIMENT 3 - Colortuning
Discussion
motion preference was assigned to the dorsal WHERE-path (Zeki, 1974; Ungerleider &
Mishkin, 1982). This strict segregation would entail the dorsal path to be non-responsive to
color and the ventral path to be non-responsive to motion. Numerous studies were
subsequently conducted, the results of which were evidently not in concordance with such a
strict segregation and receive support from results presented in this thesis regarding human
visual extra-striate regions. The following two sections discuss extra-striate regions
representing the dorsal (MT+ and V3A) and ventral path (hV4), which all show substantial
contributions to their response profile from visual attributes not specifically assigned to the
path each region is assumed to solely represent. In line with existing macaque data we find
clear color responses in human motion selective cortex MT+ (Figure 5.7), representing the
dorsal path, as well as clear luminance motion signals in ventral path region hV4 (Figure 5.5),
which is not in line with a strict segregation of ventral and dorsal path (Ungerleider &
Mishkin, 1982; Mishkin et al., 1983).
5.4.3.1 MT+
Area MT+ is believed to receive mainly magnocellular input preferring luminance
stimuli (Maunsell et al., 1990). We found that responses in MT+ elicited through chromatic
stimulation, such as L-M (parvocellular) or pure S-cone stimuli (koniocellular), presented
with sufficiently high contrast, can equal luminance responses in that area. Given the fact that
up until the first input projection sites to primary visual cortex, chromatic and luminance
signals remain fairly segregated, by means of parvo- and magnocellular projection paths
respectively (Chatterjee & Callaway, 2003), this finding suggests a cortical fusion of these
paths and indicates that color opponent information, originating in the parvocellular stream is
available to MT+ in humans as reported in monkeys (Nassi & Callaway, 2006).
Support for contributions to MT+ responses other than of magnocellular origin comes
from a recent functional imaging study, emphasizing S-cone signals in MT+ (Wandell et al.,
1999). The authors report L-M responses to be only slightly less pronounced than L+M
responses without giving exact contrast values constituting their L-M measurements (Wandell
et al., 1999). At a cone-contrast value of 10% we would reach the same conclusion (Figure
5.11), while comparisons at contrasts below 10% RMS-cone-contrast would clearly favor
luminance responses in MT+ (Figure 5.7). This observation is in concordance with results of a
single-unit investigation in macaque MT, reporting higher contrast gain for luminance as
compared to chromatic stimuli, while spiking was roughly equal at contrast levels as high as
10% for both luminance and chromatic stimulation (Gegenfurtner et al., 1994).
110
EXPERIMENT 3 - Colortuning
Discussion
Furthermore, chromatic responses in motion sensitive region MT+ seem to
interdepend more strongly on the stimulus contrast and velocity combination than luminance
responses. While for all color directions measured, lower speeds elicited generally higher
responses in human MT+ for all contrast levels, the response difference between high and low
speeds is much more pronounced for chromatic targets at low contrast levels (Figure 5.7).
This finding may indicate a speed independent processing for luminance targets in human
MT+, while responses to chromatic targets may depend more on velocity at low contrast
levels as compared to luminance signals.
5.4.3.2 V3A & hV4
A theoretical constraint, derived from psychophysical measurements on detection
thresholds regarding the direction of stimulus motion, associated with the chromatic slow
velocity mechanism of motion perception, entails a higher sensitivity to slowly moving,
chromatic stimuli in visual areas other than MT+. The restriction is imposed on MT+, because
its neural sensitivity, as assessed with single unit recordings, cannot explain the excellent
chromatic contrast motion detection and identification thresholds observed in humans and
monkeys (Gegenfurtner & Hawken, 1995, 1996). The hV4 and V3A BOLD-response profiles
measured in experiment 3 suggest an association of these two regions to the mechanism of
slow chromatic motion perception.
Cone-contrast response measurements in experiment 3 show, that while MT+
responses appear to be relatively speed invariant for luminance-emphasizing stimuli (Figure
5.7) dorsal area V3A (Figure 5.6) and ventral area hV4 (Figure 5.7) showed marked speed
dependence across all color directions tested. Low contrasts at low speeds elicit higher
BOLD-signals compared to low contrasts at high speeds as expected from observations in
Experiment 2.
As stated above and in section 5.1, the critical criterion in order to associate a visual
area to the slow channel of motion perception, this region must exhibit higher chromatic
contrast sensitivity as MT+, because MT+ single-cell responses cannot be held responsible for
chromatic detection and discrimination thresholds for slowly moving targets (Metha et al.,
1994; Gegenfurtner & Hawken, 1995; Stromeyer et al., 1995). We must thus compare
chromatic responses at low contrast levels in MT+ to candidate regions. The relative
responses to low contrast L-M stimuli in both areas (V3A: Figure 5.6, hV4: Figure 5.5, 0.3%
and 1% RMS-cone-contrast), exceeds responses at the same contrast levels in MT+ (Figure
5.7), especially at low velocities. The same effect becomes clear, comparing low contrast
responses for an additional purely chromatic color-direction. Low contrast S-cone responses
111
EXPERIMENT 3 - Colortuning
Discussion
(2.85% & 9.5% RMS-cone-contrast) for visual area V3A and hV4 exceed responses to the
same stimulus combinations in MT+. Hence, extending the stimulus range into the contrast
domain, further strengthens the conclusions made in Experiment 2.
MT+ is highly likely to account for the fast motion mechanism because it is responsive
to high stimulus velocities at the lowest luminance contrasts while also responding to
chromatic stimulation when contrast is sufficiently high. Low chromatic contrast sensitivity is
clearly observed for visual areas V3A and hV4, while the responses are enhanced for slow
moving targets as opposed to fast moving stimuli. These observations make it highly likely
that V3A and hV4 mediate the chromatic slow velocity mechanism.
5.4.4 Striate vs. extra-striate cortex
Comparing response patterns in striate and extra-striate areas reveals differences as
well as similarities. Cone-contrast stimuli theoretically stimulating the two chromatic corticogeniculate mechanisms while keeping the luminance mechanism stimulation constant, elicited
highest BOLD-responses in V1 at low and high speeds. These ‘mixed colors’ continue to
generate highest responses in extra-striate cortex as well. The reasons for this effect remain
speculative as was discussed in section 5.4.2.
The most differential response patterns between striate and extra-striate cortex are
observed as responses to stimuli emphasizing geniculo-cortical pathways are compared to
‘mixed’ colors. In V1 responses to S, L+M, L+M+S or even S-(L+M) cone-contrast stimuli
elicit the weakest response strength irrespective of stimulus velocity. This observation
underlines that cardinal mechanisms are still prevalent in V1. This changes in extra-striate
visual areas hV4 and V3A. No particular color seems to be substantially preferred over other
color combinations at low temporal frequencies, while at high temporal frequencies, cardinal
color responses drop considerably. We speculate that conscious chromatic motion perception
may be tight to activity in these two regions, because their motion sensitivity profile may
correspond to perceptual desaturation of fast moving colors.
However, these observations suggest a strong common response characteristic
between V3A and hV4 as described in Experiment 2. Strong, color indifferent BOLDresponses at low temporal frequencies support the argument that both visual areas are
involved in the processing of slowly moving chromatic stimuli. Further support to this
hypothesis is given by the fact that low contrast chromatically opponent stimuli elicited
relatively stronger responses in V3A and hV4 as compared to MT+, while the human medial
temporal area responds relatively stronger to low contrast achromatic stimuli as compared to
low contrast opponent colors.
112
EXPERIMENT 3 - Colortuning
Discussion
It seems thus feasible to conclude that the exquisite detection and motion
discrimination sensitivity to R/G-stimuli (Lee & Stromeyer, 1989) is mediated by visual areas
such as hV4 and V3A, because at very low chromatic contrasts MT+ tends to either respond
sparsely as shown in experiment 3 in human subjects or not at all, shown in
neurophysiological studies using monkeys (Gegenfurtner & Hawken, 1995, 1996). MT+
however exhibits properties making it highly likely to be involved in the fast motion channel
because it does respond well to opponent color stimuli (Gegenfurtner & Hawken, 1995;
Thiele et al., 1999; Wandell et al., 1999) and in addition, responses do not decline for high
temporal frequency moving patterns, at which responses in hV4 and V3A drop considerably
at low contrasts.
How the response profiles in MT+, V3A and hV4 may depend on activity in V1
cannot unequivocally be determined on the bases of the presented data set. It is highly likely
that the V1 response profile has a strong influence on these regions (Ferrera et al., 1994;
Wandell, 1995; Nassi & Callaway, 2006). We speculate that stimulus combinations
containing additional stimulus attributes, such as depth cues, form or moving objects, in for
instance natural scenes, may yield a slightly different response profile in V1 as observed in
this thesis, which could also lead to changes in the response profiles in hierarchically higher
regions as well. Suggestions to investigate this hypothesis are provided in the next section.
113
GENERAL DISCUSSION
Discussion
6 GENERAL DISCUSSION
Three functional magnetic resonance imaging experiments investigated the precise
nature of differential responses in human visual areas to varying stimulus combination of
color, contrast and velocity. Experiment 1 securely defined the borders of cortical visual areas
by means of the phase encoding approach and applying functional localizer measurements to
define motion sensitive regions. Experiment 2 focused on differential speed dependencies in
various visual areas by presenting radial gratings with equal cone-contrast, selectively
emphasizing psychophysically determined chromatic and achromatic detection mechanisms
and rotating at three different speeds. MT+ could be associated to a psychophysically derived
fast channel for motion perception, because it responds to chromatic stimuli and responses do
not decline at fast moving stimulus pattern. Experiment 3 investigated cone-contrast
sensitivity of human visual areas by measuring responses to stimuli modulated along thirteen
directions in cone-contrast space, parameterized in four contrast levels and presented at two
different speeds. Visual regions hV4 and V3A were associated with a psychophysically
derived slow channel for motion perception, because both regions prefer slow stimulus
velocities especially at low contrasts and chromatic sensitivity is higher as in MT+. Speed
preference in primary visual cortex was found to be strongly contrast dependent, because the
preferences change from low to high temporal frequencies as cone-contrast is increased.
The results point towards a widely distributed network of cortical regions sensitive to
stimuli defined by a combination of color and motion, while no stimulus property is the
exclusive domain of either visual area. Hence, and in concordance with newer studies
(Gegenfurtner & Hawken, 1996; Thiele et al., 1999; Wandell et al., 1999; Liu & Wandell,
2005) color and motion information seem to undergo a convergence rather than a strict
segregation as the widely held view of a ventral- (color) and dorsal- (motion) pathways would
predict (Ungerleider & Mishkin, 1982; Hubel & Livingstone, 1987). We observed a rather
gradual difference between response properties across visual areas pertaining to these
important visual attributes. Without a doubt, substantial chromatic information processing is
found in dorsal areas while a ventral area was clearly involved in processing motion stimuli.
The results are well in line with existing physiological reports in monkeys (Maunsell et al.,
1990; Nealey et al., 1991; Ferrera et al., 1994; Thiele et al., 1999; Conway & Livingstone,
2006; Nassi & Callaway, 2006; Nassi et al., 2006) and complement results from human
imaging data regarding striate and extra-striate color and motion processing (Engel et al.,
1997b; Wandell et al., 1999; Liu & Wandell, 2005).
114
GENERAL DISCUSSION
Discussion
It was found, that visual areas do exhibit preferred stimulus combinations which led us
to assign distinct visual areas to psychophysically defined motion channels (Gegenfurtner &
Hawken, 1996). The fast motion channel is believed to exhibit band-pass temporal frequency
tuning and some color selectivity while the slow motion channel is extremely color sensitive
and exhibits a low-pass temporal frequency tuning (Gegenfurtner et al., 1994; Gegenfurtner &
Hawken, 1995, 1996). Experiment 2 demonstrated temporal frequency tuning curves for area
hV4 and V3A. Strong responses were found for these regions up to 4Hz, which declined at
higher velocities (10Hz). This low-pass property makes it highly likely that neurons in V3A
and hV4 are tuned to process slowly moving chromatic stimuli. Contrarily, responses in area
MT+ on the temporal cortical surface did not decline substantially at higher speeds, preferred
medium velocities and responded well to chromatic stimuli. These properties make MT+
highly likely to be responsible for fast moving stimuli defined by either color or luminance
contrast. Furthermore, it has been proposed that neurons in monkeys MT are not sensitive
enough to account for the exquisite chromatic contrast sensitivity for slowly moving
chromatic gratings (Gegenfurtner et al., 1994; Gegenfurtner & Hawken, 1996) Our finding
that BOLD responses in V3A and hV4 exceed MT+ responses to low chromatic contrast
patterns moving at 1Hz is in concordance with and supports the view that both areas, V3A
and hV4 are more sensitive to chromatic motion than MT+ and can thus be held responsible
for the slow channel of motion perception derived psychophysically (Gegenfurtner &
Hawken, 1996).
Moreover, responses to stimuli comprising thirteen cone-contrast colors were
measured in several regions of human visual cortex, giving insight into transformation
processing from primary cortex to extra-striate areas. Responses to stimuli emphasizing
detection mechanisms (cardinal directions) are substantially reduced in primary visual cortex
as compared to ‘mixed colors’, with the exception of the R/G-mechanism (L-M), generating
quite substantial signals in V1 as well. As the visual signal is passed on to extra-striate cortex,
particularly to dorsal area V3A and ventral area hV4 these differences vanish for slow moving
colors. Both areas respond to all presented slowly moving stimuli roughly equal at 10% conecontrast and must thus apply some form of amplification pertaining to signals from cardinal
directions as they are forwarded from V1. This finding suggests that cardinal mechanisms are
not as prevalent in the neuronal population of higher tier visual areas as compared to V1. At
high speeds however, signals are reduced in both, V3A and hV4, especially for achromatic
stimuli, emphasizing their preference for slowly moving chromatic stimuli. The effect of
velocity on MT+ BOLD-responses however is less pronounced. The difference between the
hV4/V3A network and the MT+ complex becomes most obvious as high-speed achromatic
115
GENERAL DISCUSSION
Discussion
responses are compared. Most likely magno-cellularly driven fast achromatic signals generate
substantially high responses in MT+, while V3A and hV4 must receive considerably less
magnocellular signals (Nealey et al., 1991; Ferrera et al., 1994) as their high-speed
achromatic signal is well below MT+ responses at equal cone-contrast level.
The conclusions of this thesis may also help to explain results from a recent report,
demonstrating that residual chromatic motion detection and discrimination in patients with
cerebral achromatopsia is indeed similar to normal observers as long as chromatic contrast is
high enough (Cavanagh et al., 1998). This report came as a surprise, because the ventral
putative color regions in these patients have been severely damaged. The authors rejected the
notion that the residual motion processing takes place in MT+ applying the argument that the
contribution of color to motion detection in these patients was much stronger than the color
responses of monkey area MT (Cavanagh et al., 1998). At the time of the study, no chromatic
motion sensitive region besides MT+ or hV4 was known, which could account for the patients
excellent motion detection and discrimination at slow speeds. In light of the data presented in
this thesis, it appears most likely that dorsal area V3A must be responsible for residual slow
moving chromatic detection and discrimination performance in achromatopsic patients. The
data presented in this thesis showed, that V3A prefers slowly moving chromatic gratings to
luminance gratings and responses increase steeply as cone-contrast is raised.
The presented results may be a basis for future experiments tackling the phenomenon
of conscious color perception. For instance the quantitative measurements of thirteen conecontrast directions in V1 resulted in a heterogeneous response pattern across chromatic
conditions, while cardinal direction stimuli responses were among the lowest at equal conecontrast. We observed a substantial amplification of these cardinal directions responses in
hV4 and V3A at low velocities leading to roughly equal responses at 10% cone-contrast.
These responses may be biased by the heterogeneous response pattern exhibited in V1, a
precursor stage of visual processing before visual information is transmitted to higher areas.
Since the V1 response profile is highly similar across subjects, probably due to extensive
measurements and the resultant high signal to noise ratio, it may be possible to reliably
compute and present cone-contrast stimulation leading to equal responses in V1 (Engel et al.,
1997a; Engel, 2005) while simultaneously measuring extra-striate cortex. By doing so, conecontrast color tuning of higher tier visual areas in humans may be measurable independent of
differential V1 activity. In addition, applied stimulus contrast may be linked to responses of
several visual areas in order to narrow down the regions associated with conscious color
perception. It is worth noting, however, that the reported response patterns for several visual
areas are doubtlessly constrained to the stimulus properties used and may very well change as
116
GENERAL DISCUSSION
Discussion
different stimulus attributes are changed, such as spatial frequency, depth, form, size, stimulus
position or attention.
A possible extension of the presented results may constitute an application of
differential methodology to overcome the shortcomings of functional magnetic resonance
imaging to further explore and understand cortical color mechanisms and its dynamics. FMRI
is limited regarding temporal resolution of brain activity. The fastest temporal sampling of
brain activity was 1.5s in Experiment 2. It is known that visual scenes are processed
extremely fast and that the first sweep of visual information enters V1 after roughly 70 ms and
may be processed in frontal regions after 150ms (Thorpe et al., 1996). It would be of interest
in what way the geniculo-cortical paths play a differential role pertaining to dynamic
processing. For that reason, photographs of natural scenes may be generated in such a way
that they selectively stimulate either path while keeping contributions from the other two
constant. Neurophysiological methods with higher temporal resolution, such as EEG or MEG
should be used to gain insight into the cortical dynamics of the processing of natural scenes
depending on geniculo-cortical pathways.
Color is one of the most conspicuous and penetrating features of our world. The
interesting part about color however is not how it looks but rather what it represents. We as
humans share, with other old world monkeys, the extraordinary gift of three receptor classes,
which in a feat of evolution, have been connected in such a way, that we can actually attribute
fluctuations in photon properties around us, to emotions. The ability to perceive colors
develops ontogenetically earlier than human memory can trace back, making us unaware of its
merits until we may tragically lose it. Color vision should not be taken for granted.
117
APPENDIX
Experiment 2 – Speedtuning
7 APPENDIX
7.1 Experiment 2 – Speedtuning
hV4
L+M
L-M
S-L-M
L+M+S
S
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
0
1.83
0.72
-0.28
1.65
0.36
-0.92
-0.05
-1.00
-0.90
-2.60
-2.84
-3.89
-7.68
-2.89
L+M
4Hz
-0.07
0
-1.53
-1.24
0.72
-1.24
-2.61
-1.45
-2.31
-2.10
-3.91
-4.12
-4.68
-9.72
-6.07
10Hz
2.96
3.95
0
-0.82
1.53
-0.23
-2.19
-0.69
-2.04
-1.46
-3.89
-5.37
-4.79
-9.48
-5.72
1Hz
-1.29
-0.96
-3.63
0
1.95
0.67
-0.41
0.28
-1.23
-0.13
-2.26
-2.50
-3.37
-6.46
-2.68
L-M
4Hz
-1.74
-2.01
-6.17
-0.39
0
-1.80
-2.68
-2.69
-4.26
-2.02
-4.08
-3.46
-3.93
-8.46
-7.02
S-L-M
10Hz
1Hz
4Hz
-0.11 0.48
-0.47
-0.07 0.44
-0.34
-4.12 -2.71 -3.96
1.19 2.00
1.68
4.16 2.53
1.57
0
0.58
-0.35
-1.41 0
-1.47
-0.69 0.66
0
-2.16 -0.79 -1.99
-0.89 0.30
-0.32
-3.22 -2.36 -3.27
-4.20 -3.47 -2.46
-4.31 -4.37 -3.02
-8.40 -13.06 -6.68
-8.14 -3.44 -4.34
10Hz
2.42
2.08
0.07
4.26
5.22
3.33
2.74
3.34
0
0.71
-1.30
-1.65
-2.59
-5.81
-2.78
1Hz
2.42
2.21
-1.74
2.02
2.86
1.70
0.94
1.63
-1.24
0
-2.31
-3.14
-4.31
-7.51
-2.79
L+M+S
4Hz 10Hz
6.81 6.10
4.63 7.33
1.01 9.89
6.62 7.35
5.07 10.60
4.86 9.41
5.08 6.70
9.38 8.52
0.71 3.30
3.48 5.06
0
3.07
-0.84 0
-2.09 -2.96
-4.86 -4.55
-1.70 -0.90
1Hz
11.74
6.24
2.52
7.36
6.80
6.56
6.70
7.91
2.05
6.08
2.11
-1.67
0
-1.39
0.85
S
4Hz
7.13
5.58
2.99
9.14
7.36
6.35
9.21
10.19
2.71
4.29
2.64
-0.50
1.44
0
2.87
MT+
Table 4.2: Paired t-test-values for all stimulus combination in hV4 and MT+
T-values for one region were computed for the comparison of all stimulus combinations. Each entry is
the t-value obtained in the comparison of BOLD response differences elicited by the stimuli along the
columns versus the reponses obtained in the conditions along the rows. T-values for area MT are
shaded blue and green respectively. For example, in area hV4 the comparison of the BOLD responses
elicited by (L-M) stimulation at 4Hz minus the BOLD-response elicited by L+M+S at 10 Hz resulted in a
t-value of 10.6. Please note that minuend and subtrahend are swapped between hV4 and MT+.
Consequently, the t-values from one brain area must be multiplied by -1 when compared to the other.
Cone-contrast was constant at 14% for all color-directions.
118
10Hz
1.80
2.07
-0.39
2.49
3.51
2.05
1.82
2.35
-0.31
0.89
-0.98
-3.12
-2.03
-3.72
0
APPENDIX
L+M
L-M
S-L-M
L+M+S
S
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
Experiment 2 – Speedtuning
L+M
4Hz
0
0
1Hz
0
10Hz
0
0
0
1Hz
0
0
0
0
L-M
4Hz
0
0
-4.64
0
0
10Hz
0
0
0
0
0
0
1Hz
0
0
0
0
0
0
0
hV4 effect (row > column)
hV4 effect (row < column)
S-L-M
4Hz
0
0
0
0
0
0
0
0
10Hz
0
0
0
0
0.96
0
0
0
0
1Hz
0
0
0
0
0
0
0
0
0
0
L+M+S
4Hz
4.21
0
0
4.37
0.98
0
2.71
6.11
0
0
0
10Hz
3.26
3.22
0
4.85
7.14
5.21
3.23
6.06
0
1.92
0
0
1Hz
7.85
1.56
0
3.99
2.87
2.25
2.34
4.89
0
1.77
0
0
0
S
4Hz
0
0
-6.49
0
0
0
0
0
-3.10
-3.22
0
0
0
0
10Hz
0
-4.00
-6.11
0
-3.51
-6.08
0
0
0
0
0
0
0
0
0
MT+ effect (row > column)
MT+ effect (row < column)
Table 4.3: Stimulus combinations that modulate either MT+ or hV4
Stimulus combinations for which a Bonferroni corrected difference was observed in either hV4 OR in
MT+ have a non-zero entry. The non-zero values were obtained by subtracting the t-values for hV4
and MT+ in the respective color and speed combination. The green-shading of the numbers indicates
a significant effect in hV4, and the blue shading an effect in MT+. For instance: a hV4 effect of 7.14 for
the stimulus combination L+M+S 10Hz and L-M 4Hz was obtained using t-values from the same
stimulus combination in both region from Table 4.2: 10.6 – (- 3.46 * -1) = 7.14. Note: the MT+ t-value
must be multiplied with –1 to mirror its value on the matrix diagonal.
L+M
L-M
S-L-M
L+M+S
S
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
4Hz
10Hz
1Hz
0
L+M
4Hz
0
0
10Hz
0
0
0
1Hz
0
0
0
0
L-M
4Hz
0
0
0
0
0
10Hz
0
0
0
0
0
0
1Hz
0
0
0
0
0
0
0
S-L-M
4Hz
0
0
0
0
0
0
0
0
10Hz
0
0
0
0
0
0
0
0
0
1Hz
0
0
0
0
0
0
0
0
0
0
L+M+S
4Hz
0
0
0
0
0
0
0
0
0
0
0
10Hz
0
0
4.52
0
0
0
0
0
0
0
0
0
1Hz
0
0
0
0
0
0
0
0
0
0
0
0
0
S
4Hz
-0.55
-4.14
0
2.68
-1.10
-2.05
-3.85
3.51
0
0
0
0
0
0
hV4 effect
MT+ effect
Table 4.4: Stimulus combinations that modulate both MT+ and hV4
significantly.
The green shading indicates a stronger significant effect in V4 than in MT+ for the respective
comparison. The blue shading indicates a stronger effect in MT+ than in hV4. The reported number is
the difference of the t-values obtained in hV4 and MT+. It reflects the differences of the effect sizes
between the visual areas. For instance: a hV4 effect of 4.52 for the stimulus combination L+M+S 10Hz
and L+M 10Hz was obtained using t-values from the same stimulus combination in both region from
Table 4.2: 9.89 – (-5.37 * -1) = 4.52. Note: the MT+ t-value must be multiplied with –1 to mirror its
value on the matrix diagonal.
119
10Hz
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
APPENDIX
Experiment 3 - Colortuning
7.2 Experiment 3 - Colortuning
7.2.1 ANOVA tables
Dn45
V1
Effects &
Interactions
VELOCITY
(VEL)
COLOR
(COL)
VEL /
COL
CONTRAST
(CONT)
VEL /
CONT
COL/
CONT
VEL / COL /
CONT
Degrees of
freedom
Fvalue
Pvalue
Degrees of
freedom
V2d
Fvalue
Pvalue
Degrees of
freedom
V2v
Fvalue
Pvalue
Degrees of
freedom
V3d
Fvalue
Pvalue
Degrees of
freedom
V3v
Fvalue
Pvalue
Degrees of
freedom
hV4
Fvalue
Pvalue
Degrees of
freedom
V3A
Fvalue
Pvalue
Degrees of
freedom
MT+
Fvalue
Pvalue
(1.00,8.00)
0.307
0.595
(1.00,8.00)
21.943
0.002
(1.00,8.00)
10.213
0.013
(1.00,8.00)
39.092
0.000
(1.00,8.00)
15.185
0.005
(1.00,8.00)
44.982
0.000
(1.00,8.00)
34.687
0.000
(1.00,8.00)
22.200
0.002
(4.30,34.37)
14.922
0.000
(5.04,40.33)
11.272
0.000
(4.58,36.66)
16.551
0.000
(4.62,36.98)
14.232
0.000
(5.09,40.72)
20.261
0.000
(4.15,33.16)
11.484
0.000
(4.64,37.13)
31.708
0.000
(4.24,33.95)
14.943
0.000
(4.48,35.87)
3.057
0.023
(4.58,36.66)
1.197,
0.329
(4.71,37.68)
2.569
0.046
(4.77,38.18)
2.134
0.085
(5.34,42.72)
3.275
0.012
(4.43,35.40)
6.401
0.000
(4.71,37.71)
3.758
0.008
(4.04,32.28)
1.939
0.127
(1.79,14.28)
503.898
0.000
(1.45,11.57)
424.999
0.000
(1.78,14.23)
332.323
0.000
(1.66,13.31)
442.042
0.000
(2.40,19.21)
409.229
0.000
(2.29,18.33)
408.858
0.000
(1.70,13.63)
807.074
0.000
(2.11,16.88)
271.385
0.000
(1.36,10.86)
13.036
0.003
(1.85,14.78)
17.645
0.000
(1.38,11.01)
15.665
0.001
(1.42,11.37)
27.334
0.000
(2.42,19.34)
29.551
0.000
(2.34,18.72)
103.309
0.000
(1.62,12.99)
35.963
0.000
(1.88,15.04)
19.924
0.000
(5.39,43.16)
2.795
0.019
(6.53,52.24)
2.141
0.059
(5.86,46.90)
2.487
0.037
(6.23,49.84)
1.955
0.088
(5.92,47.39)
2.529
0.034
(6.34,50.74)
3.413
0.000
(6.08,48.63)
4.038
0.002
(5.60,44.83)
3.475
0.008
(6.03,48.27)
1.403
0.234
(5.47,43.79)
1.389
0.244
(6.41,51.28)
2.077
0.068
(5.75,46.04)
1.603
0.170
(6.54,52.34)
2.800
0.017
(6.21,49.66)
3.035
0.012
(6.07,48.58)
2.666
0.025
(6.43,51.44)
1.837
0.106
Lk13
V1
Effects &
Interactions
VELOCITY
(VEL)
COLOR
(COL)
VEL /
COL
CONTRAST
(CONT)
VEL /
CONT
COL/
CONT
VEL / COL /
CONT
Degrees of
freedom
Fvalue
Pvalue
Degrees of
freedom
V2d
F - value
Pvalue
Degrees of
freedom
V2v
F - value
Pvalue
Degrees of
freedom
V3d
F - value
Pvalue
Degrees of
freedom
V3v
Fvalue
Pvalue
Degrees of
freedom
hV4
Fvalue
Pvalue
Degrees of
freedom
V3A
Fvalue
Pvalue
Degrees of
freedom
MT+
Fvalue
Pvalue
(1.00,8.00)
0.002
0.969
(1.00,8.00)
81.942
0.000
(1.00,8.00)
4.647
0.063
(1.00,8.00)
63.867
0.000
(1.00,8.00)
28.239
0.001
(1.00,8.00)
24.242
0.001
(1.00,8.00)
21.069
0.002
(1.00,8.00)
4.251
0.073
(4.30,34.37)
52.968
0.000
(4.60,36.78)
59.455
0.000
(3.80,30.39)
56.759
0.000
(4.67,37.32)
89.203
0.000
(3.46,27.66)
98.718
0.000
(3.38,27.06)
35.064
0.000
(4.17,33.34)
73.845
0.000
(4.65,37.23)
71.619
0.000
(4.48,35.87)
1.543
0.206
(4.60,36.81)
4.352
0.004
(3.64,29.08)
1.296
0.295
(4.51,36.05)
3.320
0.017
(3.58,28.64)
1.650
0.193
(3.41,27.30)
1.340
0.281
(4.11,32.89)
3.136
0.026
(4.28,34.22)
1.370
0.264
(1.79,14.28)
724.294
0.000
(2.35,18.79)
1286.135
0.000
(1.85,14.78)
1112.761
0.000
(2.03,16.22)
1276.569
0.000
(1.57,12.56)
882.251
0.000
(1.46,11.64)
324.957
0.000
(1.15,9.23)
320.163
0.000
(1.58,12.68)
568.219
0.000
(1.36,10.86)
29.074
0.000
(1.60,12.84)
28.607
0.000
(1.46,11.68)
15.529
0.001
(1.29,10.28)
16.045
0.002
(1.44,11.54)
5.288
0.031
(1.19,9.55)
5.769
0.034
(1.13,9.05)
7.631
0.020
(1.62,12.95)
5.568
0.023
(5.39,43.16)
5.419
0.000
(5.99,47.89)
9.944
0.000
(5.58,44.66)
5.949
0.000
(6.15,49.18)
16.170
0.000
(5.91,47.26)
14.262
0.000
(5.49,43.91)
6.619
0.000
(6.02,48.15)
15.103
0.000
(5.55,44.42)
9.990
0.000
(6.03,48.27)
1.947
0.092
(5.37,42.98)
3.338
0.011
(5.68,45.46)
2.336
0.050
(4.75,37.97)
4.325
0.004
(6.08,48.63)
3.734
0.004
(6.08,48.62)
2.201
0.058
(4.97,39.74)
4.451
0.003
(5.96,47.65)
2.238
0.056
Table 5.1: Eight 3-way-repeated measures ANOVAs for subjects Dn45 and Lk13 for 8 ROIs.
Factors: speed (2 levels), color (13 levels) and contrast (4 levels). Note: Each speed/color/contrast combination contains 18 measurements from 9 different days
and 2 hemispheres. Degrees of freedom are Greenhouse-Geisser corrected when appropriate. Colors code uncorrected significance values: p < 0.05, p< 0.01,
p< 0.001. The critical p-value after Bonferroni correction for alpha level of 0.05 is 0.006.
120
APPENDIX
Experiment 3 - Colortuning
7.2.2 T-test tables
Table 5.2: Velocity differences for subjects Dn45 and Lk13.
T-tests were computed for each ROI at every color/contrast combination to assess speed differences
between 1Hz and 10Hz BOLD-responses as an exploratory measure of effect size. Each test
compared 18 measurements obtained at 9 separate days for 1Hz and 10Hz respectively and 2
hemispheres. Colors code uncorrected significance levels: p < 0.05, p < 0.01, p<0.001. The critical pvalue after Bonferroni correction for alpha level of 0.05 is 0.001 for a single ROI and 0.0001 over all
ROIs. Note: a negative t-value in the t-test-tables indicates a higher response for 10Hz than 1Hz for
the respective stimulus combination.
Dn45:
ConeContrast
0.14000
0.04200
L
0.01400
0.00420
0.14600
0.04380
M
0.01460
0.00440
0.95000
0.28500
S
0.09500
0.02850
0.96167
0.28850
L+M
0.09617
0.02885
0.10182
0.03055
L-M
0.01018
0.00311
0.17678
0.05303
L+S
0.01768
0.00523
0.21920
0.06576
-L+S
0.02192
0.00651
0.23052
0.06916
M+S
0.02305
0.00693
0.18950
0.05685
-M+S
0.01895
0.00566
V1
-3.17535
0.00587
1.39711
0.18146
2.39934
0.02896
3.48826
0.00304
-1.18658
0.25271
0.96743
0.34773
0.38953
0.70202
1.58254
0.13309
-1.42046
0.17467
-0.93178
0.36530
1.38888
0.18390
2.35904
0.03137
-1.23517
0.23460
-1.47732
0.15900
-1.86496
0.08063
2.74787
0.01430
-0.79226
0.43980
2.82574
0.01218
1.50576
0.15161
4.31779
0.00053
-1.84019
0.08437
0.65989
0.51871
2.96393
0.00914
2.61254
0.01885
-2.89983
0.01044
-0.38585
0.70469
0.62568
0.54035
3.26939
0.00482
-1.78143
0.09383
1.51307
0.14976
2.11873
0.05011
1.39087
0.18331
-2.04068
0.05814
3.31972
0.00434
2.15718
0.04654
2.05980
0.05607
V2d
-1.88219
0.07813
2.28311
0.03643
4.32687
0.00052
4.12128
0.00080
0.63706
0.53310
4.13559
0.00078
1.74966
0.09933
2.31134
0.03447
0.86591
0.39934
1.83284
0.08550
3.24417
0.00508
4.63173
0.00028
0.81287
0.42823
0.09774
0.92335
0.97896
0.34217
3.45577
0.00325
0.60339
0.55471
3.07611
0.00723
1.92242
0.07254
4.70619
0.00024
-0.09187
0.92794
3.55831
0.00262
3.29992
0.00452
3.34402
0.00412
-1.08966
0.29200
1.02788
0.31929
1.57847
0.13402
7.72468
0.00000
-0.37574
0.71205
4.57207
0.00031
3.82025
0.00151
2.41302
0.02818
-0.96674
0.34806
4.18821
0.00070
2.04855
0.05728
2.41248
0.02821
V2v
-1.20126
0.24713
2.05002
0.05712
2.67690
0.01654
3.07526
0.00725
-1.53554
0.14419
2.42944
0.02727
2.79646
0.01294
3.48861
0.00304
-1.79808
0.09106
1.93972
0.07025
2.54513
0.02161
4.69291
0.00024
-0.35219
0.72929
-2.63799
0.01790
0.36807
0.71764
4.24593
0.00062
-0.34656
0.73344
2.16608
0.04575
2.56913
0.02059
6.30127
0.00001
-0.57926
0.57048
3.02965
0.00797
3.35661
0.00401
2.38475
0.02981
-1.88504
0.07772
0.80139
0.43465
2.00933
0.06168
4.15714
0.00074
-1.19072
0.25113
2.93779
0.00965
3.78992
0.00161
2.53573
0.02203
-2.30863
0.03465
2.07840
0.05412
3.16680
0.00598
2.71053
0.01544
V3d
-0.40556
0.69044
3.64458
0.00218
5.91836
0.00002
4.53538
0.00034
1.76738
0.09623
5.46991
0.00005
3.64747
0.00217
4.99418
0.00013
1.91939
0.07295
2.08730
0.05321
5.46934
0.00005
5.63434
0.00004
0.78129
0.44604
0.90438
0.37921
2.41184
0.02825
4.70657
0.00024
1.76972
0.09583
4.33452
0.00051
4.30741
0.00054
5.49458
0.00005
0.76565
0.45503
4.57319
0.00031
4.99789
0.00013
5.99042
0.00002
-0.24844
0.80695
1.90982
0.07425
3.29170
0.00460
8.19264
0.00000
-0.01803
0.98584
4.41124
0.00044
6.21886
0.00001
5.88882
0.00002
0.28530
0.77907
7.00813
0.00000
3.44540
0.00333
6.94790
0.00000
V3v
-0.12622
0.90113
2.94515
0.00951
3.38354
0.00379
3.24595
0.00506
0.76792
0.45372
3.82147
0.00150
2.36203
0.03118
3.16093
0.00605
-0.18311
0.85701
1.70097
0.10830
2.51018
0.02319
4.04425
0.00094
0.80810
0.43088
-0.89831
0.38234
1.24225
0.23204
5.81014
0.00003
1.10729
0.28454
3.40734
0.00360
3.75634
0.00172
5.04555
0.00012
0.59047
0.56313
6.49834
0.00001
4.55935
0.00032
3.67306
0.00206
-1.47597
0.15936
0.93983
0.36128
3.26054
0.00491
5.05690
0.00012
-0.41226
0.68562
4.20482
0.00067
4.44794
0.00040
3.27561
0.00476
-0.17929
0.85996
2.99400
0.00859
3.16558
0.00600
4.96755
0.00014
hV4
-0.12362
0.90316
4.60759
0.00029
7.87866
0.00000
8.68555
0.00000
0.62116
0.54324
2.92378
0.00994
3.57553
0.00253
5.62396
0.00004
-0.18052
0.85901
1.18444
0.25354
7.47127
0.00000
7.42280
0.00000
0.44244
0.66409
0.51083
0.61645
0.40629
0.68992
7.86133
0.00000
0.50031
0.62367
3.31803
0.00435
8.93246
0.00000
9.56723
0.00000
0.26648
0.79327
5.52145
0.00005
6.85255
0.00000
7.16108
0.00000
-0.26882
0.79150
0.28964
0.77581
5.24070
0.00008
11.43742
0.00000
-0.21619
0.83157
5.22633
0.00008
7.61660
0.00000
5.77464
0.00003
0.96981
0.34657
5.75376
0.00003
6.53844
0.00001
10.43003
0.00000
V3A
0.63455
0.53469
3.91242
0.00124
6.62708
0.00001
5.50148
0.00005
2.70665
0.01556
4.26247
0.00060
5.41429
0.00006
6.66876
0.00001
1.96100
0.06752
3.99575
0.00104
5.17153
0.00009
4.90676
0.00016
1.62777
0.12310
1.96864
0.06657
2.39740
0.02907
5.67429
0.00003
1.20967
0.24398
6.90993
0.00000
6.09472
0.00002
7.11679
0.00000
0.97942
0.34195
6.14885
0.00001
6.15314
0.00001
5.88948
0.00002
0.86713
0.39869
2.54517
0.02161
5.73864
0.00003
9.51817
0.00000
0.70448
0.49126
4.08158
0.00087
6.17770
0.00001
5.52763
0.00005
1.23303
0.23538
8.26903
0.00000
5.29144
0.00007
10.16379
0.00000
MT-Post
1.27060
0.22204
1.85648
0.08190
5.20296
0.00009
4.94139
0.00015
0.73691
0.47184
3.35822
0.00400
3.13094
0.00645
2.81480
0.01246
0.99174
0.33609
1.91012
0.07421
1.83082
0.08582
5.32162
0.00007
2.26271
0.03792
1.52440
0.14693
0.01192
0.99064
2.71221
0.01538
-0.92935
0.36652
3.27360
0.00478
2.26750
0.03757
6.21389
0.00001
1.01561
0.32493
3.06057
0.00747
4.22335
0.00065
4.63579
0.00027
0.34076
0.73772
1.62770
0.12312
2.54471
0.02163
9.56245
0.00000
-0.32447
0.74978
2.68264
0.01634
3.87276
0.00135
2.23965
0.03967
1.34048
0.19882
4.82155
0.00019
2.71856
0.01518
5.36117
0.00006
stats
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
121
APPENDIX
1.71473
0.51442
L+M+S
0.17147
0.05144
0.11951
0.03585
L-M+S
0.01195
0.00364
0.12124
0.03637
L-M-S
0.01212
0.00364
0.90067
0.27020
S-(L+M)
0.09007
0.02702
ConeContrast
Experiment 3 - Colortuning
-2.13481
0.04859
-0.73946
0.47034
0.37596
0.71188
0.12565
0.90157
-1.38812
0.18413
0.50512
0.62036
1.59165
0.13102
0.43166
0.67175
-2.17099
0.04532
-0.47372
0.64210
2.61163
0.01889
0.48688
0.63295
-3.26185
0.00490
-0.73305
0.47413
0.12959
0.89851
3.29761
0.00454
V1
-0.59385
0.56091
0.96312
0.34982
2.54894
0.02145
1.27394
0.22088
1.42492
0.17340
1.93790
0.07049
2.21957
0.04125
0.89398
0.38458
-0.77837
0.44771
1.92811
0.07178
3.72222
0.00185
2.05601
0.05648
-0.78439
0.44427
0.45207
0.65729
2.30612
0.03482
4.09147
0.00085
V2d
-1.67277
0.11381
-0.99172
0.33610
2.13155
0.04890
1.45433
0.16519
0.14548
0.88615
1.33763
0.19972
4.33883
0.00051
0.50585
0.61986
-1.29545
0.21354
1.21657
0.24141
6.18759
0.00001
1.77514
0.09490
-1.69208
0.11001
-1.47129
0.16061
2.11892
0.05009
5.45477
0.00005
V2v
-0.49453
0.62765
0.99733
0.33345
3.47462
0.00313
3.10733
0.00678
1.88335
0.07796
4.05747
0.00091
5.18848
0.00009
1.80794
0.08945
-0.38630
0.70436
4.19246
0.00069
5.70064
0.00003
3.75334
0.00174
-0.42446
0.67688
0.78155
0.44589
3.43448
0.00340
6.90693
0.00000
V3d
-0.75823
0.45934
0.73674
0.47195
1.79830
0.09102
1.92178
0.07263
2.19493
0.04327
2.63254
0.01810
4.49964
0.00036
0.28980
0.77569
0.41896
0.68082
3.25723
0.00495
6.59679
0.00001
2.32990
0.03323
-1.23981
0.23292
-0.68786
0.50139
2.22485
0.04083
7.65344
0.00000
V3v
-0.59996
0.55693
0.75459
0.46146
2.99781
0.00852
3.13069
0.00645
1.74633
0.09992
3.35540
0.00402
8.33213
0.00000
2.93285
0.00975
0.13350
0.89546
4.63929
0.00027
8.36058
0.00000
5.85330
0.00002
-0.95683
0.35289
0.70222
0.49263
2.99850
0.00851
6.54154
0.00001
hV4
-0.26448
0.79479
1.25789
0.22648
3.46681
0.00318
3.63153
0.00224
2.95818
0.00925
5.89920
0.00002
7.55887
0.00000
2.46941
0.02517
0.52697
0.60544
4.50618
0.00036
5.96538
0.00002
5.42291
0.00006
-0.56118
0.58245
1.19080
0.25110
3.17294
0.00590
10.71512
0.00000
V3A
-0.28159
0.78187
0.63052
0.53726
2.67424
0.01663
1.53833
0.14351
1.50528
0.15174
2.74559
0.01436
4.46850
0.00039
0.78140
0.44597
-1.26844
0.22279
2.54318
0.02170
5.66973
0.00003
3.05969
0.00749
0.50302
0.62180
0.83050
0.41848
1.81953
0.08759
5.87609
0.00002
MT-Post
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
stats
Lk13:
ConeContrast
0.14000
0.04200
L
0.01400
0.00420
0.14600
0.04380
M
0.01460
0.00440
0.95000
0.28500
S
0.09500
0.02850
0.96167
0.28850
L+M
0.09617
0.02885
0.10182
0.03055
L-M
0.01018
0.00311
0.17678
0.05303
L+S
0.01768
0.00523
0.21920
0.06576
-L+S
0.02192
0.00651
V1
-4.34885
0.00050
-0.68084
0.50570
2.90115
0.01042
1.77768
0.09447
-2.73421
0.01470
1.42300
0.17394
1.62486
0.12373
0.35001
0.73090
-2.10103
0.05184
0.54559
0.59287
2.82419
0.01222
3.54872
0.00267
-4.03389
0.00096
-0.35302
0.72868
-0.97358
0.34476
4.43470
0.00042
-3.33425
0.00420
0.49608
0.62658
3.95275
0.00114
4.13674
0.00077
-4.70341
0.00024
1.77233
0.09538
3.24157
0.00511
1.44340
0.16820
-3.79787
0.00158
-0.50381
0.62126
1.75091
0.09911
2.42251
0.02765
V2d
-1.60692
0.12763
2.05125
0.05699
9.03125
0.00000
3.18173
0.00580
1.92631
0.07202
4.61366
0.00029
6.30917
0.00001
4.87692
0.00017
1.02928
0.31866
3.85349
0.00140
6.75344
0.00000
6.16024
0.00001
-0.56573
0.57942
1.63024
0.12258
1.02410
0.32102
6.17671
0.00001
0.50178
0.62266
6.28808
0.00001
8.86520
0.00000
7.94521
0.00000
-1.11840
0.27991
4.70164
0.00024
5.12292
0.00010
3.33963
0.00416
-0.65213
0.52358
1.66374
0.11562
8.19861
0.00000
5.89075
0.00002
V2v
-2.35918
0.03136
0.80631
0.43189
3.50375
0.00294
0.62667
0.53971
-0.06954
0.94542
1.86735
0.08028
2.27390
0.03710
1.15454
0.26523
-1.29624
0.21328
1.79668
0.09129
2.90398
0.01036
3.44851
0.00330
-2.51709
0.02287
0.13997
0.89043
0.36489
0.71997
6.26244
0.00001
-1.25530
0.22740
2.41069
0.02831
3.16259
0.00603
2.08601
0.05335
-3.18157
0.00580
3.73525
0.00180
3.61850
0.00231
1.39650
0.18164
-2.56666
0.02069
0.51975
0.61035
2.66610
0.01690
3.57518
0.00253
V3d
-0.03022
0.97627
2.38154
0.03000
10.22987
0.00000
3.06554
0.00739
2.85571
0.01144
6.04322
0.00002
6.80459
0.00000
5.61491
0.00004
2.18016
0.04452
7.65096
0.00000
7.68485
0.00000
6.40930
0.00001
1.74802
0.09962
2.08717
0.05323
2.09045
0.05290
7.57094
0.00000
0.15924
0.87548
6.60290
0.00001
9.95568
0.00000
7.42442
0.00000
-0.35418
0.72783
7.47637
0.00000
5.49177
0.00005
3.91083
0.00125
1.04643
0.31091
2.29274
0.03575
10.92019
0.00000
6.08234
0.00002
V3v
0.49519
0.62720
2.05126
0.05699
6.21170
0.00001
1.82688
0.08643
3.50279
0.00295
4.19689
0.00068
4.27072
0.00059
4.81193
0.00019
1.99626
0.06321
3.64303
0.00219
4.69871
0.00024
5.68526
0.00003
0.81746
0.42567
1.65501
0.11740
2.86768
0.01116
9.20537
0.00000
1.35205
0.19516
4.83221
0.00018
5.73645
0.00003
5.19463
0.00009
-0.12259
0.90396
5.57982
0.00004
4.88656
0.00016
1.60075
0.12899
1.18424
0.25361
1.54492
0.14191
6.39994
0.00001
5.45614
0.00005
hV4
0.37221
0.71462
3.38760
0.00376
5.66874
0.00003
2.06947
0.05505
1.43833
0.16961
2.94894
0.00943
3.59563
0.00242
3.98936
0.00106
1.05478
0.30719
3.92673
0.00120
4.35614
0.00049
5.98924
0.00002
1.25898
0.22610
1.16144
0.26250
1.90527
0.07488
8.09973
0.00000
0.44993
0.65879
4.02811
0.00097
5.99219
0.00002
4.20254
0.00067
-0.42743
0.67477
4.23047
0.00064
5.33753
0.00007
2.24128
0.03954
1.16939
0.25937
1.06289
0.30361
6.51947
0.00001
5.51411
0.00005
V3A
-0.23796
0.81493
1.70334
0.10785
9.01321
0.00000
2.44086
0.02666
1.36596
0.19085
5.66307
0.00004
5.24183
0.00008
5.07667
0.00011
1.40540
0.17902
5.63060
0.00004
5.11773
0.00010
4.63409
0.00028
1.93579
0.07077
1.97127
0.06624
1.07109
0.30002
7.08669
0.00000
-0.57468
0.57350
4.72457
0.00023
8.05334
0.00000
5.04928
0.00012
-0.74071
0.46960
5.11407
0.00010
5.19937
0.00009
2.87948
0.01090
1.19057
0.25119
1.67283
0.11380
6.96299
0.00000
7.97172
0.00000
MT-Post
-0.35428
0.72775
-0.21563
0.83200
4.31926
0.00053
0.62592
0.54019
0.25767
0.79995
2.62090
0.01854
1.38957
0.18369
1.48736
0.15636
0.69494
0.49706
4.01140
0.00101
1.75494
0.09840
1.90068
0.07551
0.73688
0.47186
1.64967
0.11850
-0.45584
0.65463
4.44745
0.00041
-2.08240
0.05371
3.08472
0.00710
3.47789
0.00311
1.06326
0.30345
-2.02804
0.05954
1.92107
0.07272
1.77999
0.09407
1.99401
0.06348
0.73219
0.47464
-0.14767
0.88445
3.23863
0.00514
4.44801
0.00040
stats
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
122
APPENDIX
0.23052
0.06916
M+S
0.02305
0.00693
0.18950
0.05685
-M+S
0.01895
0.00566
1.71473
0.51442
L+M+S
0.17147
0.05144
0.11951
0.03585
L-M+S
0.01195
0.00364
0.12124
0.03637
L-M-S
0.01212
0.00364
0.90067
0.27020
S-(L+M)
0.09007
0.02702
ConeContrast
Experiment 3 - Colortuning
-2.56021
0.02097
2.50266
0.02355
4.71453
0.00023
1.25158
0.22871
-1.13080
0.27481
-0.86062
0.40216
4.02127
0.00099
2.63212
0.01812
-2.80758
0.01264
-3.04388
0.00774
0.62187
0.54279
1.75865
0.09774
-3.40870
0.00359
2.58042
0.02012
1.23748
0.23376
1.13282
0.27398
-2.90512
0.01033
0.07476
0.94133
2.61814
0.01864
1.04223
0.31280
-4.17653
0.00071
-1.34582
0.19712
1.41369
0.17662
3.33112
0.00423
V1
1.67987
0.11240
4.50890
0.00036
10.45838
0.00000
3.78047
0.00164
0.97923
0.34204
3.79025
0.00161
5.90575
0.00002
6.97292
0.00000
-1.29566
0.21347
-0.51661
0.61250
3.10776
0.00677
5.08981
0.00011
-0.67515
0.50922
4.78561
0.00020
4.80494
0.00019
2.97926
0.00885
0.13081
0.89755
4.03637
0.00096
7.57967
0.00000
2.13878
0.04822
-0.71704
0.48368
2.07911
0.05405
2.86134
0.01131
8.12432
0.00000
V2d
0.38245
0.70716
3.42493
0.00347
6.45903
0.00001
1.32892
0.20251
-0.03800
0.97016
0.73305
0.47413
4.57046
0.00031
2.03966
0.05825
-1.92212
0.07258
-0.12291
0.90371
2.54563
0.02159
2.50745
0.02332
-1.35282
0.19492
3.05562
0.00755
1.53074
0.14536
2.29712
0.03545
-0.77491
0.44969
0.66324
0.51662
2.50079
0.02364
-0.31252
0.75868
-2.73692
0.01462
-0.86743
0.39853
2.36645
0.03091
4.27605
0.00058
V2v
3.16271
0.00603
5.59896
0.00004
13.47046
0.00000
3.25400
0.00498
2.33994
0.03258
7.31365
0.00000
7.69018
0.00000
6.34601
0.00001
0.41029
0.68703
0.57311
0.57454
4.22291
0.00065
6.00999
0.00002
0.69215
0.49876
5.30551
0.00007
6.80484
0.00000
4.04247
0.00094
1.06489
0.30273
4.82578
0.00019
7.72712
0.00000
3.38270
0.00380
0.50097
0.62322
3.26938
0.00482
2.99919
0.00849
9.72698
0.00000
V3d
2.26411
0.03782
4.43019
0.00042
8.75611
0.00000
3.25574
0.00496
1.99076
0.06387
3.31344
0.00439
4.79879
0.00020
4.95469
0.00014
0.72356
0.47978
1.58964
0.13148
3.74141
0.00178
4.67393
0.00025
1.51953
0.14814
5.63440
0.00004
4.28794
0.00056
3.34763
0.00409
1.52856
0.14590
3.01510
0.00822
5.53930
0.00004
0.95537
0.35360
0.97464
0.34424
2.05125
0.05699
3.58409
0.00248
6.31764
0.00001
V3v
1.50768
0.15113
4.11805
0.00081
8.91712
0.00000
3.86914
0.00136
1.20637
0.24521
3.46306
0.00320
5.53996
0.00004
5.11214
0.00010
0.97060
0.34619
1.05790
0.30581
3.23090
0.00523
4.08930
0.00086
0.77115
0.45186
3.27934
0.00472
5.73768
0.00003
3.36959
0.00390
0.94727
0.35759
2.34086
0.03252
6.40921
0.00001
1.39332
0.18258
0.86535
0.39964
1.64948
0.11854
2.86138
0.01131
5.83097
0.00003
hV4
1.12005
0.27922
4.83666
0.00018
8.99270
0.00000
2.43413
0.02702
1.66380
0.11561
4.28962
0.00056
7.97840
0.00000
6.11496
0.00001
0.58885
0.56418
0.14508
0.88646
1.53307
0.14479
4.08601
0.00086
0.44459
0.66257
4.68379
0.00025
7.50762
0.00000
3.31466
0.00438
0.24105
0.81258
3.88137
0.00132
6.06094
0.00002
2.73277
0.01475
0.51634
0.61268
1.85534
0.08207
2.37090
0.03064
6.52420
0.00001
V3A
-0.04184
0.96714
2.29421
0.03565
3.57684
0.00252
-0.27713
0.78523
0.53946
0.59700
0.90617
0.37829
2.15657
0.04660
1.47881
0.15861
0.20916
0.83696
-0.73371
0.47374
1.03307
0.31693
1.64753
0.11895
-1.83656
0.08492
2.69261
0.01601
1.58781
0.13189
1.74098
0.10088
-0.96685
0.34801
1.04451
0.31178
3.94496
0.00116
0.21326
0.83382
0.36530
0.71967
1.19489
0.24954
0.94097
0.36071
1.96049
0.06759
MT-Post
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
T-value
p-Value
stats
123
CITATIONS
8 CITATIONS
ALBRECHT, D. G. (1995). Visual cortex neurons in monkey and cat: effect of
contrast on the spatial and temporal phase transfer functions. Vis Neurosci 12,
1191-1210.
ALBRECHT, D. G. & HAMILTON, D. B. (1982). Striate cortex of monkey and cat:
contrast response function. J Neurophysiol 48, 217-237.
ALITTO, H. J. & USREY, W. M. (2004). Influence of contrast on orientation and
temporal frequency tuning in ferret primary visual cortex. J Neurophysiol 91,
2797-2808.
BANDETTINI, P. A., JESMANOWICZ, A., WONG, E. C. & HYDE, J. S. (1993).
Processing strategies for time-course data sets in functional MRI of the human
brain. Magn Reson Med 30, 161-173.
BARTELS, A. & ZEKI, S. (2000). The architecture of the colour centre in the
human visual brain: new results and a review. Eur J Neurosci 12, 172-193.
BEAUCHAMP, M. S., HAXBY, J. V., JENNINGS, J. E. & DEYOE, E. A. (1999). An
fMRI version of the Farnsworth-Munsell 100-Hue test reveals multiple colorselective areas in human ventral occipitotemporal cortex. Cereb Cortex 9,
257-263.
BECK, D. M. & KASTNER, S. (2005). Stimulus context modulates competition in
human extrastriate cortex. Nat Neurosci. 8, 1110-1116. .
BENARDETE, E. A. & KAPLAN, E. (1999). The dynamics of primate M retinal
ganglion cells. Vis Neurosci 16, 355-368.
BENARDETE, E. A., KAPLAN, E. & KNIGHT, B. W. (1992 ). Contrast gain control in
the primate retina: P cells are not X-like, some M cells are. Vis Neurosci 8,
483-486.
BETZOLD , W. V. (1873). Über das Gesetz der Farbenmischung und die
physiologischen Grundfarben. Annalen der Physiologischen Chemie 226, 221
- 247.
BETZOLD , W. V. (1874). Die Farbenlehre in Hinblick auf Kunst und
Kunstgewerbe, Braunschweig.
BOYNTON, G. M., ENGEL, S. A., GLOVER, G. H. & HEEGER, D. J. (1996). Linear
systems analysis of functional magnetic resonance imaging in human V1. J
Neurosci 16, 4207-4221.
BRAINARD, D. H., ROORDA, A., YAMAUCHI, Y., CALDERONE, J. B., METHA, A.,
NEITZ, M., NEITZ, J., WILLIAMS, D. R. & GH., J. (2000). Functional
consequences of the relative numbers of L and M cones. J Opt Soc Am A 17,
607-614.
124
CITATIONS
BREWER, A. A., LIU, J., WADE, A. R. & WANDELL, B. A. (2005). Visual field maps
and stimulus selectivity in human ventral occipital cortex. Nat Neurosci. 8,
1102-1109. Epub 2005 Jul 1117.
BRIGGS, F. & CALLAWAY, E. M. (2005 ). Laminar patterns of local excitatory
input to layer 5 neurons in macaque primary visual cortex. Cereb Cortex 15,
479-488.
BROWN, P. K. & WALD, G. (1964). Visual pigments in single rods and cones of
human retina. Science 144, 45-52.
BUCKNER, R. L., BANDETTINI, P. A., O'CRAVEN, K. M., SAVOY, R. L., PETERSEN,
S. E., RAICHLE, M. E. & ROSEN, B. R. (1996). Detection of cortical activation
during averaged single trials of a cognitive task using functional magnetic
resonance imaging. Proc Natl Acad Sci 93, 14878-14883.
CALKINS, D. J., TSUKAMOTO, Y., STERLING, P.,. (1996). Foveal cones form basal
as well as invaginating junctions with diffuse ON bipolar cells. Vision Res 36,
3373-3381.
CALLAWAY, E. M. (2004). Feedforward, feedback and inhibitory connections in
primate visual cortex. Neural Netw 17, 625-632.
CALLAWAY, E. M. (2005). Neural substrates within primary visual cortex for
interactions between parallel visual pathways. Prog Brain Res 149, 59-64.
CARANDINI, M., HEEGER, D. J. & MOVSHON, J. A. (1997). Linearity and
normalization in simple cells of the macaque primary visual cortex. 17 21.
CASAGRANDE, V. A. (1994). A third parallel visual pathway to primate area V1.
Trends Neurosci 17, 305-310.
CAVANAGH, P. & ANSTIS, S. (1991). The contribution of color to motion in
normal and color-deficient observers. Vision Res 31, 2109-2148.
CAVANAGH, P., HENAFF, M. A., MICHEL, F., LANDIS, T., TROSCIANKO, T. &
INTRILIGATOR, J. (1998). Complete sparing of high-contrast color input to
motion perception in cortical color blindness. Nat Neurosci. 1, 242-247.
CAVANAGH, P., MACLEOD, D. I. & ANSTIS, S. M. (1987 ). Equiluminance: spatial
and temporal factors and the contribution of blue-sensitive cones. J Opt Soc
Am A 4, 1428-1438.
CAVANAGH, P., TYLER, C. W. & FAVREAU, O. E. (1984). Perceived velocity of
moving chromatic gratings. J Opt Soc Am A. 1, 893-899.
CHARLES, E. R. & LOGOTHETIS, N. K. (1989). The responses of middle temporal
(MT) neurons to isoluminant stimuli. Invest Ophthalmol Vis Sci.
125
CITATIONS
CHATTERJEE, S. & CALLAWAY, E. M. (2003). Parallel colour-opponent pathways
to primary visual cortex. Nature 426, 668-671.
CHATTERJEE, S., CALLAWAY, E.M.,. (2002). S cone contributions to the
magnocellular visual pathway in macaque monkey. Neuron 35, 1135-1146.
CIE. (1926). Commision Internationale de l'eclairage Proceedings,1924.
Cambridge University Press, Cambridge UK, 1926.
COLE, G. R., HINE, T. & MCILHAGGA, W. (1993). Detection mechanisms in L-,
M-, and S-cone contrast space. J Opt Soc Am A. 10, 38-51.
CONWAY, B. R. (2001). Spatial structure of cone inputs to color cells in alert
macaque primary visual cortex (V1). J Neurosci. 21, 2768-2783.
CONWAY, B. R., HUBEL, D. H. & LIVINGSTONE, M. S. (2002). Color contrast in
macaque V1. Cereb Cortex 12, 915-925.
CONWAY, B. R. & LIVINGSTONE, M. S. (2003). The neural basis for color vision.
In Encyclopedia of Cognitive Science. ed. NADEL, L., pp. 568-576. Nature
Publishing Group, London.
CONWAY, B. R. & LIVINGSTONE, M. S. (2006). Spatial and temporal properties of
cone signals in alert macaque primary visual cortex. J Neurosci 26, 1082610846.
CORBETTA, M., MIEZIN, F. M., DOBMEYER, S., SHULMAN, G. L. & PETERSEN, S. E.
(1991). Selective and divided attention during visual discriminations of shape,
color, and speed: functional anatomy by positron emission tomography. J
Neurosci 11, 2383-2402.
CRONER, L. J. & ALBRIGHT, T. D. (1999). Segmentation by color influences
responses of motion-sensitive neurons in the cortical middle temporal visual
area. Journal of Neuroscience 19, 3935-3951.
DACEY, D. M. (1993). Morphology of a small-field bistratified ganglion cell type
in the macaque and human retina. Vis Neurosci 10, 1081-1098.
DACEY, D. M. (2000). Parallel pathways for spectral coding in primate retina.
Annu. Rev. Neurosci. 23.
DACEY, D. M. & LEE, B. B. (1994). The 'blue-on' opponent pathway in primate
retina originates from a distinct bistratified ganglion cell type. Nature 367, 731735.
DACEY, D. M., PETERSON, B. B., ROBINSON, F. R., GAMLIN, P. D.,. (2003).
Fireworks in the primate retina. In vitro photodynamics reveals diverse LGNprojecting ganglion cell types. Neuron 37 15-27.
DALE, A. M. (1999). Optimal experimental design for event-related fMRI.
Human Brain Mapping 8, 109-114.
126
CITATIONS
DALE, A. M. & BUCKNER, R. L. (1997). Selective averaging of rapidly presented
individual trials using fMRI. Human Brain Mapping 5, 329-340.
DANIEL, P. M. & WHITTERIDGE, D. (1961). The representation of the visual field
on the cerebral cortex in monkeys. J Physiol (Paris) 159, 203-221.
DAW, N. W. (1968). Colour-coded ganglion cells in the goldfish retina:
extension of their receptive fields by means of new stimuli. J Physiol 197, 567592.
DE MONASTERIO, F. M., SCHEIN, S. J. (1982). Spectral bandwidths of coloropponent cells of geniculocortical pathway of macaque monkeys. J
Neurophysiol 47, 214-224.
DE VALOIS, R. L., ABRAMOV, I. & JACOBS, G. H. (1966). Analysis of response
patterns of LGN cells. J Opt Soc Am 56, 966-977.
DE WEERD, P., PERALTA, M. R., 3RD, DESIMONE, R. & UNGERLEIDER, L. G.
(1999). Loss of attentional stimulus selection after extrastriate cortical lesions
in macaques. Nat Neurosci 2, 753-758.
DEAN, A. F. & TOLHURST, D. J. (1986). Factors influencing the temporal phase
of response to bar and grating stimuli for simple cells in the cat striate cortex.
Exp Brain Res 62, 143-151.
DERRINGTON, A. M. & HENNING, G. B. (1993). Detecting and discriminating the
direction of motion of luminance and colour gratings. Vision Res. 33, 799-811.
DERRINGTON, A. M., KRAUSKOPF, J. & LENNIE, P. (1984). Chromatic
mechanisms in lateral geniculate nucleus of macaque. J Physiol 357, 241-265.
DESIMONE, R. & SCHEIN, S. J. (1987). Visual properties of neurons in area V4
of the macaque: sensitivity to stimulus form. J Neurophysiol. 57, 835-868.
DEYOE, E. A., BANDETTINI, P., NEITZ, J., MILLER, D. & WINANS, P. (1994).
Functional magnetic resonance imaging (FMRI) of the human brain. J
Neurosci Methods 54, 171-187.
DEYOE, E. A. & VAN ESSEN, D. C. (1985). Segregation of efferent connections
and receptive field properties in visual area V2 of the macaque. Nature 317,
58-61.
DEYOE, E. A. & VAN ESSEN, D. C. (1988). Concurrent processing streams in
monkey visual cortex. Trends Neurosci 11, 219-226.
DOBKINS, K. R. & ALBRIGHT, T. D. (1994). What happens if it changes color
when it moves?: the nature of chromatic input to macaque visual area MT. J
Neurosci. 14, 4854-4870.
127
CITATIONS
DOBKINS, K. R., GUNTHER, K. L. & PETERZELL, D. H. (2000). What covariance
mechanisms underlie green/red equiluminance, luminance contrast sensitivity
and chromatic (green/red) contrast sensitivity? Vision Res 40, 613-628.
DOBKINS, K. R., THIELE, A. & ALBRIGHT, T. D. (2000a). Comparison of redgreen equiluminance points in humans and macaques: evidence for different
L:M cone ratios between species. J Opt Soc Am A Opt Image Sci Vis. 17, 545556.
DOUGHERTY, R. F., KOCH, V. M., BREWER, A. A., FISCHER, B., MODERSITZKI, J. &
WANDELL, B. A. (2003). Visual field representations and locations of visual
areas V1/2/3 in human visual cortex. Journal of Vision 3, 586-598.
DOUGHERTY, R. F., PRESS, W. A. & WANDELL, B. A. (1999). Perceived speed of
colored stimuli. Neuron. 24, 893-899.
DOW, B. M. & GOURAS, P. (1973). Color and spatial specificity of single units in
Rhesus monkey foveal striate cortex. J Neurophysiol 36, 79-100.
DUMOULIN, S. O., BITTAR, R. G., KABANI, N. J., BAKER, C. L., JR., LE GOUALHER,
G., BRUCE PIKE, G. & EVANS, A. C. (2000). A new anatomical landmark for
reliable identification of human area V5/MT: a quantitative analysis of sulcal
patterning. Cereb Cortex. 10, 454-463.
EJIMA, Y., TAKAHASHI, S. (1984). Bezold-Brucke hue shift and nonlinearity in
opponent-color process. Vision Research 24, 1897-1904.
ENGEL, S. A. (2005). Adaptation of oriented and unoriented color-selective
neurons in human visual areas. Neuron 45, 613-623.
ENGEL, S. A. & FURMANSKI, C. S. (2001). Selective adaptation to color contrast
in human primary visual cortex. J Neurosci 21, 3949-3954.
ENGEL, S. A., GLOVER, G. H. & WANDELL, B. A. (1997a). Retinotopic
organization in human visual cortex and the spatial precision of functional
MRI. Cereb Cortex 7, 181-192.
ENGEL, S. A., RUMELHART, D. E., WANDELL, B. A., LEE, A. T., GLOVER, G. H.,
CHICHILNISKY, E. J. & SHADLEN, M. N. (1994). fMRI of human visual cortex.
Nature 369, 525.
ENGEL, S. A., ZHANG, X. & WANDELL, B. (1997b). Colour tuning in human visual
cortex measured with functional magnetic resonance imaging. Nature. 388,
68-71.
ESKEW, R. T., MCLELLAN, J. S. & GIULIANINI, F. (1999). Chromatic detection
and discrimination. In Color Vision. ed. GEGENFURTNER, K. R. & SHARPE, L. T.,
pp. 345-368. Cambridge University Press, Cambridge.
128
CITATIONS
FELLEMAN, D. J. & VAN ESSEN, D. C. (1987). Receptive field properties of
neurons in area V3 of macaque monkey extrastriate cortex. J Neurophysiol
57, 889-920.
FELLEMAN, D. J. & VAN ESSEN, D. C. (1991). Distributed hierarchical
processing in the primate cerebral cortex. Cereb Cortex 1, 1-47.
FELLEMAN, D. J., XIAO, Y. & MCCLENDON, E. (1997a). Modular organization of
occipito-temporal pathways: cortical connections between visual area 4 and
visual area 2 and posterior inferotemporal ventral area in macaque monkeys.
J Neurosci 17, 3185-3200.
FERRERA, V. P., NEALEY, T. A. & MAUNSELL, J. H. (1994). Responses in
macaque visual area V4 following inactivation of the parvocellular and
magnocellular LGN pathways. J Neurosci, 4.
FITZPATRICK, D. (1996). The functional organization of local circuits in visual
cortex: insights from the study of tree shrew striate cortex. Cereb Cortex 6,
329-341.
FITZPATRICK, D., LUND, J. S. & BLASDEL, G. G. (1985). Intrinsic connections of
macaque striate cortex: afferent and efferent connections of lamina 4C. J
Neurosci 5, 3329-3349.
FIZE, D., VANDUFFEL, W., NELISSEN, K., DENYS, K., CHEF D'HOTEL C.,
FAUGERAS, O. & ORBAN, G. A. (2003). The retinotopic organization of primate
dorsal V4 and surrounding areas: A functional magnetic resonance imaging
study in awake monkeys. J Neurosci 23, 7395-7406.
FOSTER, K. H., GASKA, J. P., NAGLER, M. & POLLEN, D. A. (1985). Spatial and
temporal frequency selectivity of neurones in visual cortical areas V1 and V2
of the macaque monkey. J Physiol 365, 331-363.
GEGENFURTNER, K. R. & HAWKEN, M. J. (1995). Temporal and chromatic
properties of motion mechanisms. Vision Res. 35, 1547-1563.
GEGENFURTNER, K. R. & HAWKEN, M. J. (1996). Interaction of motion and color
in the visual pathways. Trends Neurosci 19, 394-401.
GEGENFURTNER, K. R. & KIPER, D., C.,. (2003). Color vision. Annu Rev
Neurosci 26, 181-206.
GEGENFURTNER, K. R., KIPER, D. C., BEUSMANS, J. M., CARANDINI, M., ZAIDI, Q.
& MOVSHON, J. A. (1994). Chromatic properties of neurons in macaque MT.
Vis Neurosci. 11, 455-466.
GEGENFURTNER, K. R., KIPER, D. C. & FENSTEMAKER, S. B. (1996). Processing
of color, form, and motion in macaque area V2. Vis Neurosci 13, 161-172.
GEGENFURTNER, K. R., KIPER, D. C. & LEVITT, J. B. (1997). Functional
properties of neurons in macaque area V3. J Neurophysiol. 77, 1906-1923.
129
CITATIONS
GEGENFURTNER, K. R. & SHARPE, L. T. (1999). Color vision: from genes to
perception. Cambridge UP, Cambridge, UK.
GEGENFURTNER, K. R., SHARPE, L.T.,. (1999). Color Vision: from genes to
perception. Cambridge University Press., Cambridge.
GHOSE, G. M. & TS'O, D. Y. (1997). Form processing modules in primate area
V4. J Neurophysiol. 77, 2191-2196.
GOODALE, M. A. & MILNER, A. D. (1992). Separate visual pathways for
perception and action. Trends Neurosci 15, 20-25.
GOURAS, P. (1974). Opponent-colour cells in different layers of foveal striate
cortex. J Physiol 238, 583-602.
GRILL-SPECTOR, K. & MALACH, R. (2001). fMR-adaptation: a tool for studying
the functional properties of human cortical neurons. Acta Psychol (Amst) 41,
1409-1422.
GRILL-SPECTOR, K., SAYRES, R. & RESS, D. (2006). High-resolution imaging
reveals highly selective nonface clusters in the fusiform face area. Nat
Neurosci 9, 1177-1185.
GRUNERT, U., GREFERATH, U., BOYCOTT, B. B., WASSLE, H. (1993). Parasol (P
alpha) ganglion-cells of the primate fovea: immunocytochemical staining with
antibodies against GABAA-receptors. Vision Res 33, 1-14.
GULYAS, B., HEYWOOD, C. A., POPPLEWELL, D. A., ROLAND, P. E. & COWEY, A.
(1994). Visual form discrimination from color or motion cues: functional
anatomy by positron emission tomography. Proc Natl Acad Sci U S A 91,
9965-9969.
GUR, M. & SNODDERLY, D. M. (1997). A dissociation between brain activity and
perception: chromatically opponent cortical neurons signal chromatic flicker
that is not perceived. Vision Res. 37, 377-382.
HADJIKHANI, N., LIU, A. K., DALE, A. M., CAVANAGH, P. & TOOTELL, R. B. (1998).
Retinotopy and color sensitivity in human visual cortical area V8. Nat Neurosci
1, 235-241.
HAGSTROM, S. A., NEITZ, J. & NEITZ, M. (1998). Variations in cone populations
for red–green color vision examined by analysis of mRNA,. NeuroReport 9,
1963-1967.
HASSON, U., HAREL, M., LEVY, I. & MALACH, R. (2003). Large-scale mirrorsymmetry organization of human occipito-temporal object areas. Neuron 37,
1027-1041.
130
CITATIONS
HASSON, U., LEVY, I., BEHRMANN, M., HENDLER, T. & MALACH, R. (2002).
Eccentricity bias as an organizing principle for human high-order object areas.
Neuron 34, 479-490.
HAWKEN, M. J., GEGENFURTNER, K. R. & TANG, C. (1994). Contrast
dependence of colour and luminance motion mechanisms in human vision.
Nature 367, 268-270.
HAWKEN, M. J., SHAPLEY, R. M. & GROSOF, D. H. (1996). Temporal-frequency
selectivity in monkey visual cortex. Vis Neurosci. 13, 477-492.
HEEGER, D. J. (1999). Linking visual perception with human brain activity. Curr
Opin Neurobiol 9, 474-479.
HELMHOLTZ, H. V. (1867/1925). Handbuch der Physiologischen Optik, vol. 1st
ed. Voss, Leipzig.
HENDRY, S. H. & YOSHIOKA, T. (1994). A neurochemically distinct third channel
in the macaque dorsal lateral
geniculate nucleus. Science 264, 575–577.
HERING, E. (1878/1964). Handbuch der gesamten Augenheilkunde.Grundzüge
der Lehre vom Lichtsinn., Berlin, 1905.
HERING, E. (1964). Outlines of a theory of the light sense. Harvard University
Press, Cambridge.
HESS, R. H., BAKER, C. L., JR. & ZIHL, J. (1989). The "motion-blind" patient:
low-level spatial and temporal filters. J Neurosci. 9, 1628-1640.
HEYWOOD, C. A. & COWEY, A. (1987). On the role of cortical area V4 in the
discrimination of hue and pattern in macaque monkeys. J Neurosci 7, 26012617.
HEYWOOD, C. A., GADOTTI, A. & COWEY, A. (1992). Cortical area V4 and its role
in the perception of color. J Neurosci 12, 4056-4065.
HICKS, T. P., LEE, B. B. & VIDYASAGAR, T. R. (1983). The responses of cells in
macaque lateral geniculate nucleus to sinusoidal gratings. J Physiol 337, 183200.
HINRICHS, H., SCHOLZ, M., TEMPELMANN, C., WOLDORFF, M. G., DALE, A. M. &
HEINZE, H. J. (2000). Deconvolution of event-related fMRI responses in fastrate experimental designs: tracking amplitude variations. J Cogn Neurosci 12,
76-89.
HOLUB, R. A. & MORTON-GIBSON, M. (1981). Response of Visual Cortical
Neurons of the cat to moving sinusoidal gratings: response-contrast functions
and spatiotemporal interactions. J Neurophysiol 46, 1244-1259.
131
CITATIONS
HORN, B. K. P. (1974). Determining lightness from an image. Computer
Graphics Image Processing 3, 111−299.
HORTON, J. C. & HOYT, W. F. (1991a). The representation of the visual field in
human striate cortex. A revision of the classic Holmes map. Arch Ophthalmol
109, 816-824.
HUBEL, D. H. & LIVINGSTONE, M. S. (1987). Segregation of form, color, and
stereopsis in primate area 18. J Neurosci 7, 3378-3415.
HUBEL, D. H. & WIESEL, T. N. (1968). Receptive fields and functional
architecture of monkey striate cortex. J Physiol 195, 215-243.
HUBEL, D. H. W. T. N. (1959). Receptive fields of single neurons in the cat's
striate Cortex. J. Physiol. (Lond.) 148, 574-591.
HUBEL, D. H. W. T. N. (1962). Receptive fields, binocular interaction, and
functional architecture of the cat's visual cortex. Journal of Physiology
(London) 160, 106-154.
HUK, A. C., DOUGHERTY, R. F. & HEEGER, D. J. (2002). Retinotopy and
functional subdivision of human areas MT and MST. J Neurosci 22, 71957205.
HUK, A. C. & HEEGER, D. J. (2000). Task-related modulation of visual cortex. J
Neurophysiol 83, 3525-3536.
HURVICH, L. M. & JAMESON, D. (1955). Some quantitative aspects of an
opponent-colors theory. II. Brightness, saturation, and hue in normal and
dichromatic vision. J Opt Soc Am 45, 602-616.
HURVICH, L. M. & JAMESON, D. (1957). An opponent process theory of color
vision. Psychological Review 64, 384-404.
HURVICH, L. M., JAMESON, D. (1956). Some quantitative aspects of an
opponent-colors theory. IV. A psychological color specification system. J Opt
Soc Am 46, 416-421.
INOUYE, T. (1909). Die Sehstöhrungen bei Schussverletzungen der kortikalen
Sehsphäre nach Beobachtungen an Verwundeten der letzten japanischen
Kriege. Engelmann, Leipzig.
JACOBY, R. A., WIECHMANN, A.F., AMARA, S.G., LEIGHTON, B.H., MARSHAK,
D,W.,. (2000). Diffuse bipolar cells provide input to OFF parasol ganglion cells
in the macaque retina. J Comp Neurol 416, 6-18.
JAMESON, D., HURVICH, L.M. (1956 ). Some quantitative aspects of an
opponent-colors theory. III. Changes in brightness, saturation, and hue with
chromatic adaptation. J Opt Soc Am 46, 405-415.
132
CITATIONS
JOHNSON, E. N., HAWKEN, M. J. & SHAPLEY, R. (2001). The spatial
transformation of color in the primary visual cortex of the macaque monkey.
Nat Neurosci 4, 409-416.
JOHNSON, E. N., HAWKEN, M. J. & SHAPLEY, R. (2004). Cone inputs in macaque
primary visual cortex. J Neurophysiol. 91, 2501-2514. Epub 2004 Jan 2528.
JOSEPHS, O., TURNER, R. & FRISTON, K. J. (1997). Event-Related fMRI. Human
Brain Mapping 5, 243-248.
KAISER, P. K. & BOYNTON, R. M. (1996). Human Color Vision. Opt. Soc. Am.,
Washington, DC.
KAUFMAN, L. (1974). Sight and Mind: An introduction to visual perception.
Oxford University Press, Oxford.
KAYSER, A., PRIEBE, N. J. & MILLER, K. D. (2001). Contrast-dependent
nonlinearities arise locally in a model of contrast-invariant orientation tuning. J
Neurophysiol 85, 2130-2149.
KIPER, D. C., FENSTEMAKER, S. B. & GEGENFURTNER, K. R. (1997). Chromatic
properties of neurons in macaque area V2. Vis Neurosci. 14, 1061-1072.
KLEINSCHMIDT, A., LEE, B. B., REQUARDT, M. & FRAHM, J. (1996). Functional
mapping of color processing by magnetic resonance imaging of responses to
selective P- and M-pathway stimulation. Exp Brain Res 110, 279-288.
KLUG, K., TUSKAMOTO, Y., STERLING, P., SCHEIN, S.J.,. (1993). Blue cone offmidget ganglion cells in macaque. Investigative Opthalmology and Visual
Science 34, 1993.
KRAUSKOPF, J. (2000a). A journey in color space. In Color Research &
Application.Special Issue: The Proceedings of the International Colour Vision
Society . Issue Edited by C. R. Cavonius, Kenneth Knoblauch, Barry B. Lee,
Joel Pokorny., vol. 26S2 - S11, pp. S2 - S11.
KRAUSKOPF, J. (2000b). Relative number of long- and middle-wavelengthsensitive cones in the human fovea. Journal of the Optical Society of America
A 17
510-516.
KRAUSKOPF, J. & FARELL, B. (1990). Influence of colour on the perception of
coherent motion. Nature. 348, 328-331.
KRAUSKOPF, J., WILLIAMS, D. R. & HEELEY, D. W. (1982). Cardinal directions of
color space. Vision Res 22, 1123-1131.
KREMERS, J., WEISS, S. & ZRENNER, E. (1997). Temporal properties of
marmoset lateral geniculate cells. Vision Res 37, 2649-2660
133
CITATIONS
KRIES VON, J. (1905). Die Gesichtsempfindungen. In W. Nagel (Ed.),
Handbuch der Physiologie der Menschen. Vieweg, Brunswick.
LACHICA, E. A., BECK, P. D. & CASAGRANDE, V. A. (1992). Parallel pathways in
macaque monkey striate cortex: anatomically defined columns in layer III.
Proc Natl Acad Sci U S A 89, 3566-3570.
LAND, E. H. (1959). Experiments in color vision. Sci Am 200, 84-94.
LAND, E. H. (1977). The retinex theory of color vision. Sci Am 237, 108-128.
LAND, E. H. & MCCANN, J. J. (1971 ). Lightness and retinex theory. J Opt Soc
Am. 61, 1-11
LANDISMAN, C. E. & TS'O, D. Y. (2002). Color Processing in Macaque Striate
Cortex: Electrophysiological Properties. The Journal of Neurophysiology 87,
3138-3151.
LEE, B. B. (1996). Receptive field structure in the primate retina. Vision Res
36, 631-644.
LEE, B. B., MARTIN, P. R. & VALBERG, A. (1988). The physiological basis of
heterochromatic flicker photometry demonstrated in the ganglion cells of the
macaque retina. J Physiol 404, 323-347.
LEE, B. B., POKORNY, J., SMITH, V. C., MARTIN, P. R., VALBERG, A. (1990).
Luminance and chromatic modulation sensitivity of macaque ganglion cells
and human observers. J Opt Soc Am A. 7, 2223-2236.
LEE, B. B., SILVEIRA, L.C., YAMADA, E., KREMERS, J.,. (1996a). Parallel
pathways in the retina of Old and New World primates. Rev Bras Biol 56 :,
323-338.
LEE, J. & STROMEYER, C. F., 3RD. (1989). Contribution of human short-wave
cones to luminance and motion detection. J Physiol 413, 563-593.
LENNIE, P., D’ZMURA, M. . (1988). Mechanisms of color vision. Crit. Rev.
Neurobiol. 3, 333-400
LENNIE, P., KRAUSKOPF, J. & SCLAR, G. (1990). Chromatic mechanisms in
striate cortex of macaque. J Neurosci 10, 649-669.
LENNIE, P., POKORNY, J. & SMITH, V. C. (1993). Luminance. J Opt Soc Am A
10, 1283-1293.
LEVINSON, E. & SEKULAR, R. (1975). The independence of channels in human
vision selective for direction of movement. J Physiol. 250(2), 347-366.
LEVITT, J. B., KIPER, D. C. & MOVSHON, J. A. (1994). Receptive fields and
functional architecture of macaque V2. J Neurophysiol 71, 2517-2542.
134
CITATIONS
LEVY, I., HASSON, U., AVIDAN, G., HENDLER, T. & MALACH. R. (2001). Centerperiphery organization of human object areas. Nat Neurosci 4, 533-539.
LINDSEY, D. T. & TELLER, D. Y. (1990). Motion at isoluminance:
discrimination/detection ratios for moving isoluminant gratings. Vision Res 30,
1751-1761.
LIU, J. & WANDELL, B. A. (2005). Specializations for chromatic and temporal
signals in human visual cortex. J Neurosci. 25, 3459-3468.
LIU, T., SLOTNICK, S. D. & YANTIS, S. (2004). Human MT+ mediates perceptual
filling-in during apparent motion. Neuroimage. 21, 1772-1780.
LIVINGSTONE, M. & HUBEL, D. (1988). Segregation of form, color, movement,
and depth: anatomy, physiology, and perception. Science 240, 740-749.
LIVINGSTONE, M. S. & HUBEL, D. H. (1984). Anatomy and physiology of a color
system in the primate visual cortex. J Neurosci 4, 309-356.
LIVINGSTONE, M. S. & HUBEL, D. H. (1987). Connections between layer 4B of
area 17 and the thick cytochrome oxidase stripes of area 18 in the squirrel
monkey. J Neurosci 7, 3371-3377.
LOGOTHETIS, N. K. (2003). The underpinnings of the BOLD functional magnetic
resonance imaging signal. J Neurosci 23, 3963-3971.
LOGOTHETIS, N. K., PAULS, J., AUGATH, M., TRINATH, T. & OELTERMANN, A.
(2001). Neurophysiological investigation of the basis of the fMRI signal. Nature
412, 150-157.
LOGOTHETIS, N. K. & WANDELL, B. A. (2004). Interpreting the BOLD signal.
Annu Rev Physiol 66, 735-769.
LUECK, C. J., ZEKI, S., FRISTON, K. J., DEIBER, M. P., COPE, P., CUNNINGHAM, V.
J., LAMMERTSMA, A. A., KENNARD, C. & FRACKOWIAK, R. S. (1989). The colour
centre in the cerebral cortex of man. Nature 340, 386-389.
LUND, J. S., WU, Q., HADINGHAM, P. T. & LEVITT, J. B. (1995). Cells and circuits
contributing to functional properties in area V1 of macaque monkey cerebral
cortex: bases for neuroanatomically realistic models. J Anat 187, 563-581.
MACLEOD, D. I. & BOYNTON, R. M. (1979). Chromaticity diagram showing cone
excitation by stimuli of equal luminance. J Opt Soc Am 69, 1183-1186.
MARKS, W. B., DOBELLE, W. H. & MACNICHOL, M. F. J. (1964). Visual pigment
of single primate cones. Science 143, 1181-1183.
MARROCCO, R. T., MCCLURKIN, J. W. & YOUNG, R. A. (1982). Spatial
summation and conduction latency classification of cells of the lateral
geniculate nucleus of macaques. J Neurosci 2, 1275-1291.
135
CITATIONS
MAUNSELL, J. H., NEALEY, T. A. & DEPRIEST, D. D. (1990). Magnocellular and
parvocellular contributions to responses in the middle temporal visual area
(MT) of the macaque monkey. J Neurosci 10, 3323-3334.
MAXWELL, J. C. (1855). Experiments on colour, as perceived by the eye, with
remarks on colour-blindness. Transactions of the Royal Society, Edinborough
21, 275-298.
MCKEEFRY, D. J. & ZEKI, S. (1997). The position and topography of the human
colour centre as revealed by functional magnetic resonance imaging. Brain
120 ( Pt 12), 2229-2242.
MEADOWS, J. C. (1974). Disturbed perception of colours associated with
localized cerebral lesions. Brain 97, 615-632.
MERIGAN, W. H. (1996). Basic visual capacities and shape discrimination after
lesions of extrastriate area V4 in macaques. Vis Neurosci 13, 51-60.
MERIGAN, W. H. (2000). Cortical area V4 is critical for certain texture
discriminations, but this effect is not dependent on attention. Vis Neurosci 17,
949-958.
MERIGAN, W. H. & MAUNSELL, J. H. (1993). How parallel are the primate visual
pathways? Annu Rev Neurosci 16, 369-402.
MERIGAN, W. H., MAUNSELL, J.H. (1993). How parallel are the primate visual
pathways? Annu Rev Neurosci 16, 369-402.
MERIGAN, W. H. & PHAM, H. A. (1998). V4 lesions in macaques affect both
single- and multiple-viewpoint shape discriminations. Vis Neurosci 15, 359367.
METHA, A. B., VINGRYS, A. J. & BADCOCK, D. R. (1994). Detection and
discrimination of moving stimuli: the effects of color, luminance, and
eccentricity. J Opt Soc Am A Opt Image Sci Vis. 11, 1697-1709.
MISHKIN, M., UNGERLEIDER, L. G. & MACKO, K. A. (1983). Object Vision and
Spatial Vision: Two cortical pathways. Visual Perception.
MORAN, J. & DESIMONE, R. (1985). Selective attention gates visual processing
in the extrastriate cortex. Science 229, 782-784.
MORRONE, M. C., DENTI, V. & SPINELLI, D. (2002). Color and luminance
contrasts attract independent attention. Curr Biol 12, 1134-1137.
MOVSHON, J. A., ADELSON, E. H., GIZZI, M. S. & NEWSOME, W. T. (1985). The
analysis of moving visual patterns. In Study Week on Pattern Recognition
Mechanisms. ed. CHAGAS, C., GATTASS, R. & GROSS, C., pp. 117-151. Vatican
Press, Rome.
136
CITATIONS
MOVSHON, J. A., THOMPSON, I. D. & TOLHURST, D. J. (1978). Spatial and
temporal contrast sensitivity of neurones in areas 17 and 18 of the cat's visual
cortex. J Physiol. 283, 101-120.
MUCKLI, L., KOHLER, A., KRIEGESKORTE, N. & SINGER, W. (2005). Primary visual
cortex activity along the apparent-motion trace reflects illusory perception.
PLoS Biol. 3, e265. Epub 2005 Jul 2019.
MULLEN, K. T. (1985). The contrast sensitivity of human colour vision to redgreen and blue-yellow chromatic gratings. J Physiol 359, 381-400.
MULLEN, K. T. & BOULTON, J. C. (1992). Absence of smooth motion perception
in color vision. Vision Res. 32, 483-488.
MULLEN, K. T., DUMOULIN, S. O., MCMAHON, K. L., BRYANT, M., DE ZUBICARAY,
G. I. & HESS, R. F. (2005). A comparison of the BOLD fMRI response to
achromatic, L/M opponent and S-cone opponent cardinal stimuli in human
visual cortex: I. perceptually matched vs contrast matched stimuli [Abstract].
Journal of Vision 5, 95a.
MÜLLER, G. E. (1924). Darstellung und Erklärung der verschiedenen Typen der
Farbenblindheit. Vandenhoed & Ruprecht, Göttingen.
NASSI, J. J. & CALLAWAY, E. M. (2006). Multiple circuits relaying primate
parallel visual pathways to the middle temporal area. J Neurosci 26, 1278912798.
NASSI, J. J., LYON, D. C. & CALLAWAY, E. M. (2006). The parvocellular LGN
provides a robust disynaptic input to the visual motion area MT. Neuron 50,
319-327.
NEALEY, T. A., FERRERA, V. P. & MAUNSELL, J. H. R. (1991). Magnocellular and
parvocellular contributions to the ventral extrastriate cortical processing
stream. Soc. neurosci. Abstr. 17, 525.
NERGER, J. L. & CICERONE, C. M. (1992). The ratio of L cones to M cones in
the human parafoveal retina. Vision Res 32, 879-888.
NEWSOME, W. T., BRITTEN, K. H., SALZMAN, C. D. & MOVSHON, J. A. (1991).
Neuronal mechanisms of motion perception. Cold Spring Harb Symp Quant
Biol 55, 697-705.
NEWSOME, W. T., WURTZ, R. H., DURSTELER, M. R. & MIKAMI, A. (1985).
Deficits in visual motion processing following ibotenic acid lesions of the
middle temporal visual area of the macaque monkey. J Neurosci. 5, 825-840.
NEWTON, I. (1704). Optiks. Smith and Walford, London.
NEWTON, J. R. & ESKEW, R. T. (2003). Chromatic detection and discrimination
in the periphery: a postreceptoral loss of color sensitivity. Vis Neurosci 20,
511-521.
137
CITATIONS
OGAWA, S., LEE, T. M., KAY, A. R. & TANK, D. W. (1990). Brain magnetic
resonance imaging with contrast dependent on blood oxygenation. Proc Natl
Acad Sci U S A 87, 9868-9872.
OGAWA, S., TANK, D. W., MENON, R., ELLERMANN, J. M., KIM, S. G., MERKLE, H.
& UGURBIL, K. (1992). Intrinsic signal changes accompanying sensory
stimulation: functional brain mapping with magnetic resonance imaging. Proc
Natl Acad Sci U S A 89, 5951-5955.
ORBAN, G. A., HOFFMANN, K. P. & DUYSENS, J. (1985). Velocity selectivity in
the cat visual system. I. Responses of LGN cells to moving bar stimuli: a
comparison with cortical areas 17 and 18. J Neurophysiol 54, 1026-1049.
PALMER, G. (1777). Theory of light, London:Leacroft.
PANIGRAHI, G. (1973). Penetrod Land Color Display System (PENTECOST)
and some observations concerning color perception, UIUCDCS-R-73-583,
Department of Computer Science, University of Illinois at Urbana-Champaign,
Urbana, Illinois, Oct. 1973.
PASUPATHY, A. & CONNOR, C. E. (2001). Shape representation in area V4:
position-specific tuning for boundary conformation. J Neurophysiol. 86, 25052519.
PAZO-ALVAREZ, P., AMENEDO, E. & CADAVEIRA, F. (2004). Automatic detection
of motion direction changes in the human brain. Eur J Neurosci 19, 19781986.
PERRY, V. H. & COWEY, A. (1985). The ganglion cell and cone distributions in
the monkey's retina: implications for central magnification factors. Vision Res
25, 1795-1810.
PERRY, V. H., OEHLER, R., COWEY, A. (1984). Retinal ganglion cells that project
to the dorsal lateral geniculate nucleus in the macaque monkey. Neuroscience
12, 1101-1123.
PETERHANS, E. & VON DER HEYDT, R. (1993). Functional organization of area
V2 in the alert macaque. Eur J Neurosci 5, 509-524.
POGGIO, G. F., BAKER, F. H., MANSFIELD, R. J., SILLITO, A. & GRIGG, P. (1975).
Spatial and chromatic properties of neurons subserving foveal and parafoveal
vision in rhesus monkey. Brain Res 100, 25-59.
PRESS, W. A., BREWER, A. A., DOUGHERTY, R. F., WADE, A. R. & WANDELL, B.
A. (2001). Visual areas and spatial summation in human visual cortex. Vision
Res 41, 1321-1332.
RAMACHANDRAN, V. S. & GREGORY, R. L. (1978). Does colour provide an input
to human motion perception? Nature 275, 55-56.
138
CITATIONS
RODIECK, R. W. (1998). The first steps in seeing. Sinauer.
ROE, A. W. & TS'O, D. Y. (1999). Specificity of color connectivity between
primate V1 and V2. J Neurophysiol 82, 2719-2730.
ROORDA, A., WILLIAMS, D. R. (1999). The arrangement of the three cone
classes in the living human eye. Nature 397, 520-522.
ROSEN, B. R., BUCKNER, R. L. & DALE, A. M. (1998). Event-related functional
MRI: past, present, and future. Proc Natl Acad Sci 95, 773-780.
SAITO, H., YUKIE, M., TANAKA, K., HIKOSAKA, K., FUKADA, Y. & IWAI, E. (1986).
Integration of direction signals of image motion in the superior temporal sulcus
of the macaque monkey. J Neurosci. 6, 145-157.
SALIN, P. A. & BULLIER, J. (1995). Corticocortical connections in the visual
system: structure and function. Physiol Rev 75, 107-154.
SALZMAN, C. D., BRITTEN, K. H. & NEWSOME, W. T. (1990). Cortical
microstimulation influences perceptual judgements of motion direction. Nature
346, 174-177.
SANKERALLI, M. J. & MULLEN, K. T. (1996). Estimation of the L-, M- and S-cone
weights of the post-receptoral detection mechanisms. Journal of the Optical
Society of America A 13, 906-915.
SCHEIN, S. J. & DESIMONE, R. (1990). Spectral properties of V4 neurons in the
macaque. J Neurosci 10, 3369-3389.
SCHEIN, S. J., MARROCCO, R. T. & DE MONASTERIO, F. M. (1982). Is there a high
concentration of color-selective cells in area V4 of monkey visual cortex? J
Neurophysiol 47, 193-213.
SCHILLER, P. H. (1993). The effects of V4 and middle temporal (MT) area
lesions on visual performance in the rhesus monkey. Vis Neurosci 10, 717746.
SCHILLER, P. H. (1995). Effect of lesions in visual cortical area V4 on the
recognition of transformed objects. Nature 376, 342-344.
SCHLUPPECK, D. & ENGEL, S. A. (2002). Color opponent neurons in V1: a
review and model reconciling results from imaging and single-unit recording. J
Vis 2, 480-492.
SCHNAPF, J. L., KRAFT, T. W. & BAYLOR, D. A. (1987). Spectral sensitivity of
human cone photoreceptors. Nature. 325, 439-441.
SEIDEMANN, E., POIRSON, A. B., WANDELL, B. A. & NEWSOME, W. T. (1999).
Color signals in area MT of the macaque monkey. Neuron. 24, 911-917.
139
CITATIONS
SERENO, M. I., DALE, A. M., REPPAS, J. B., KWONG, K. K., BELLIVEAU, J. W.,
BRADY, T. J., ROSEN, B. R. & TOOTELL, R. B. (1995). Borders of multiple visual
areas in humans revealed by functional magnetic resonance imaging.
Science. 268, 889-893.
SHADLEN, M. N., BRITTEN, K. H., NEWSOME, W. T. & MOVSHON, J. A. (1996). A
computational analysis of the relationship between neuronal and behavioral
responses to visual motion. J Neurosci 16, 1486-1510.
SHAPLEY, R. M. & VICTOR, J. D. (1978). The effect of contrast on the transfer
properties of cat retinal ganglion cells. 285.
SHAPLEY, R. M. & VICTOR, J. D. (1981). How the contrast gain control modifies
the frequency responses of cat retinal ganglion cells. J Physiol 318, 161-179.
SHARPE, L. T. & NORDBY, K. (1990). The photoreceptors in the achromat. In:.
In Night Vision: Basic, Clinical and Applied Aspects. ed. HESS RF, S., L.T.,
NORDBY, K., ED, pp. 335-389. Cambridge University Press, Cambridge.
SHIPP, S., DE JONG, B. M., ZIHL, J., FRACKOWIAK, R. S. & ZEKI, S. (1994). The
brain activity related to residual motion vision in a patient with bilateral lesions
of V5. Brain. 117, 1023-1038.
SINCICH, L. C. & HORTON, J. C. (2005). The circuitry of V1 and V2: integration
of color, form, and motion. Annu Rev Neurosci 28, 303-326.
SLOAN, K. R. J. & BAJCSY, R. (1975). A computational structure for color
perception. Proceedings of the 1975 annual conference, 42 - 45.
SMITH, A. T., GREENLEE, M. W., SINGH, K. D., KRAEMER, F. M. & HENNIG, J.
(1998). The processing of first- and second-order motion in human visual
cortex assessed by functional magnetic resonance imaging (fMRI). J
Neurosci. 18, 3816-3830.
SMITH, A. T., SINGH, K. D., WILLIAMS, A. L. & GREENLEE, M. W. (2001).
Estimating receptive field size from fMRI data in human striate and extrastriate
visual cortex. Cereb Cortex 11, 1182-1190.
SMITH, V. C., POKORNY, J., DAVIS, M. & YEH. T. (1994). Mechanisms
subserving temporal modulation sensitivity in silent-cone substitution. J Opt
Soc Am A Opt Image Sci Vis 12, 241-249.
SOLOMON, S. G. & LENNIE, P. (2005). Chromatic gain controls in visual cortical
neurons. J Neurosci 25, 4779-4792.
SPILLMANN, L. & WERNER, J., EDITORS. (1990). Visual Perception: The
neurophysiological foundations. Academic Press.
STOCKMAN, A. & SHARPE, L. T. (1998). Human cone spectral sensitivities: a
progress report. Vision Res. 38, 3193-3206.
140
CITATIONS
STOCKMAN, A. & SHARPE, L. T. (2000). The spectral sensitivities of the middleand long-wavelength-sensitive cones derived from measurements in
observers of known genotype. Vision Res 40, 1711-1737.
STOCKMAN, A., SHARPE, L. T. & FACH, C. (1999). The spectral sensitivity of the
human short-wavelength sensitive cones derived from thresholds and color
matches. Vision Res 39, 2901-2927.
STROMEYER, C. F., 3RD, KRONAUER, R. E., RYU, A., CHAPARRO, A. & ESKEW, R.
T., JR. (1995). Contributions of human long-wave and middle-wave cones to
motion detection. J Physiol. 485, 221-243.
TALBOT, S. & MARSHALL, W. (1941). Physiological studies on neural
mechanisms of visual localization and discrimination. Am J Ophthalmol 24,
1255-1263.
TEO, P. C., SAPIRO, G. & WANDELL, B. A. (1997). Creating connected
representations of cortical gray matter for functional MRI visualization. IEEE
Trans Med Imaging. 16, 852-863.
THIELE, A., DOBKINS, K. R. & ALBRIGHT, T. D. (1999). The contribution of color
to motion processing in Macaque middle temporal area. J Neurosci. 19, 65716587.
THORELL, L. G., DE VALOIS, R. L. & ALBRECHT, D. G. (1984). Spatial mapping of
monkey V1 cells with pure color and luminance stimuli. Vision Res 24, 751769.
THORPE, S., FIZE, D. & MARLOT, C. (1996). Speed of processing in the human
visual system. Nature 381, 520 - 522.
THULBORN, K. R., WATERTON, J. C., MATTHEWS, P. M. & RADDA, G. K. (1982).
Oxygenation dependence of the transverse relaxation time of water protons in
whole blood at high field. Biochim Biophys Acta 714, 265-270.
TOLIAS, A. S., KELIRIS, G. A., SMIRNAKIS, S. M. & LOGOTHETIS, N. K. (2005a).
Neurons in macaque area V4 acquire directional tuning after adaptation to
motion stimuli. Nat Neurosci 8, 591-593.
TOLIAS, A. S., KELIRIS, G. A., SMIRNAKIS, S. M. & LOGOTHETIS, N. K. (2005b).
Neurons in macaque area V4 acquire directional tuning after adaptation to
motion stimuli. Nat Neurosci. 8, 591-593. Epub 2005 Apr 2017.
TOLIAS, A. S., MOORE, T., SMIRNAKIS, S. M., TEHOVNIK, E. J., SIAPAS, A. G. &
SCHILLER, P. H. (2001a). Eye movements modulate visual receptive fields of
V4 neurons. Neuron. 29, 757-767.
TOLIAS, A. S., SMIRNAKIS, S. M., AUGATH, M. A., TRINATH, T. & LOGOTHETIS, N.
K. (2001b). Motion processing in the macaque: revisited with functional
magnetic resonance imaging. J Neurosci. 21, 8594-8601.
141
CITATIONS
TOOTELL, R. B., HADJIKHANI, N., HALL, E. K., MARRETT, S., VANDUFFEL, W.,
VAUGHAN, J. T. & DALE, A. M. (1998). The retinotopy of visual spatial attention.
Neuron 21, 1409-1422.
TOOTELL, R. B., HADJIKHANI N.,. (2001). Where is 'dorsal V4' in human visual
cortex? Retinotopic, topographic and functional evidence. Cereb Cortex 11,
298-311.
TOOTELL, R. B., MENDOLA, J. D., HADJIKHANI, N. K., LEDDEN, P. J., LIU, A. K.,
REPPAS, J. B., SERENO, M. I. & DALE, A. M. (1997). Functional analysis of V3A
and related areas in human visual cortex. J Neurosci 17, 7060-7078.
TOOTELL, R. B., REPPAS, J. B., DALE, A. M., LOOK, R. B., SERENO, M. I.,
MALACH, R., BRADY, T. J. & ROSEN, B. R. (1995a). Visual motion aftereffect in
human cortical area MT revealed by functional magnetic resonance imaging.
Nature. 375, 139-141.
TOOTELL, R. B., REPPAS, J. B., KWONG, K. K., MALACH, R., BORN, R. T., BRADY,
T. J., ROSEN, B. R. & BELLIVEAU, J. W. (1995b). Functional analysis of human
MT and related visual cortical areas using magnetic resonance imaging. J
Neurosci. 15, 3215-3230.
TS'O, D. Y. & GILBERT, C. D. (1988). The organization of chromatic and spatial
interactions in the primate striate cortex. J Neurosci. 8, 1712-1727.
TURNER, R., LE BIHAN, D., MOONEN, C. T., DESPRES, D. & FRANK, J. (1991).
Echo-planar time course MRI of cat brain oxygenation changes. Magn Reson
Med 22, 159-166.
TYLER, C. W., LIKOVA, L. T., CHEN, C. C., KONTSEVICH, L. L., SCHIRA, M. M. &
WADE, A. R. (2005). Extended Concepts of Occipital Retinotopy. Current
Medical Imaging Reviews 1, 319-329(311).
UNGERLEIDER, L. D. & MISHKIN, M. (1982). Two cortical visual systems. In:
Analysis of visual behavior. MIT Press, Ingle, D.J., Goodale. M.D. & Mansfield,
R.J. (Ed.), Cambridge.
USREY, W. M. & REID, R. C. (2000). Visual physiology of the lateral geniculate
nucleus in two species of new world monkey: Saimiri sciureus and Aotus
trivirgatis. J Physiol 523 755-769.
VAN ESSEN, D. C., ANDERSON C.H. (1990). Information processing strategies
and pathways in the primate retina and visual cortex. In Introduction to Neural
and Electronic Networks. ed. S.F. ZORNETZER, J. L. D., C. LAU, pp. 43-72.
Academic Press, Orlando.
VAN ESSEN, D. C., ANDERSON, C. H. & FELLEMAN, D. J. (1992). Information
processing in the primate visual system: an integrated systems perspective.
Science 255, 419-423.
142
CITATIONS
VAN ESSEN, D. C. & DEYOE, E. A. (1995). Concurrent processing in the primate
visual cortex. In The cognitive neuroscience. ed. GAZZANIGA, M. S., pp. 383400. MIT Press, Cambridge,MA.
VAN ESSEN, D. C. & GALLANT, J. L. (1994). Neural mechanisms of form and
motion processing in the primate visual system. Neuron 13, 1-10.
WACHTLER, T., SEJNOWSKI, T. J. & ALBRIGHT, T. D. (2003). Representation of
color stimuli in awake macaque primary visual cortex. Neuron 37, 681-691.
WADE, A. R., BREWER, A. A., RIEGER, J. W. & WANDELL, B. A. (2002).
Functional measurements of human ventral occipital cortex: retinotopy and
colour. Philos Trans R Soc Lond B Biol Sci 357, 963-973.
WALSH, V., BUTLER, S. R., CARDEN, D. & KULIKOWSKI, J. J. (1992). The effects
of V4 lesions on the visual abilities of macaques: shape discrimination. Behav
Brain Res 50, 115-126.
WALSH, V., LE MARE, C., BLAIMIRE, A. & COWEY, A. (2000). Normal
discrimination performance accompanied by priming deficits in monkeys with
V4 or TEO lesions. NeuroReport 11, 1459-1462.
WANDELL, B. (1995). Foundations of vision. Sinauer, Sunderland,
Massachusetts.
WANDELL, B. A. (1999). Computational neuroimaging of human visual cortex.
Annu Rev Neurosci 22, 145-173.
WANDELL, B. A. (2004). Computational neuroimaging: Cortical color signals.
Biophysical Journal 86, 515a-515a.
WANDELL, B. A., BREWER, A. A. & DOUGHERTY, R. F. (2005). Visual field map
clusters in human cortex. Philos Trans R Soc Lond B Biol Sci 360, 693-707.
Review.
WANDELL, B. A., POIRSON, A. B., NEWSOME, W. T., BASELER, H. A., BOYNTON,
G. M., HUK, A., GANDHI, S. & SHARPE, L. T. (1999). Color signals in human
motion-selective cortex. Neuron 24, 901-909.
WÄSSLE, H., GRUNERT, U., ROHRENBECK, J. & BOYCOTT, B. B. (1990). Retinal
ganglion cell density and cortical magnification factor in the primate. Vision
Res 30, 1897-1911.
WATSON, A. B. & ROBSON, J. G. (1981). Discrimination at threshold: labelled
detectors in human vision. Vision Res 21, 1115-1122.
WATSON, A. B., THOMPSON, P. G., MURPHY, B. J. & NACHMIAS, J. (1980).
Summation and discrimination of gratings moving in opposite directions.
Vision Res 20, 341-347.
143
CITATIONS
WATSON, J. D., MYERS, R., FRACKOWIAK, R. S., HAJNAL, J. V., WOODS, R. P.,
MAZZIOTTA, J. C., SHIPP, S. & ZEKI, S. (1993). Area V5 of the human brain:
evidence from a combined study using positron emission tomography and
magnetic resonance imaging. Cereb Cortex. 3, 79-94.
WEBSTER, M. A. & MOLLON, J. D. (1993). Contrast adaptation dissociates
different measures of luminous efficiency. J Opt Soc Am A 10, 1332-1340.
WEBSTER, M. A. & MOLLON, J. D. (1997). Motion minima for different directions
in color space. Vision Res 37, 1479-1498.
WIESEL, T. N. & HUBEL, D. H. (1966). Spatial and chromatic interactions in the
lateral geniculate body of the rhesus monkey. J Neurophysiol. 29, 1115-1156.
WIKLER, K. C., WILLIAMS, R. W. & RAKIC, P. (2004). Photoreceptor mosaic:
Number and distribution of rods and cones in the rhesus monkey retina. The
Journal of Comparative Neurology 297, 499 - 508.
WYSZECKI, G. & STILES, W. S. (1982). Color Science Concepts and Methods,
Quantitative Data and Formulae, second edition. John Wiley, New York.
YANTIS, S. (2005). How visual salience wins the battle for awareness. Nat
Neurosci. 8, 975-977.
YEH, T., LEE, B. B. & KREMERS, J. (1995a). Temporal response of ganglion
cells of the macaque retina to cone-specific modulation. J Opt Soc Am A Opt
Image Sci Vis 12, 456-464.
YEH, T., LEE, B. B., KREMERS, J., COWING, J. A., HUNT, D. M., MARTIN, P. R. &
TROY, J. B. (1995). Visual responses in the lateral geniculate nucleus of
dichromatic and trichromatic marmosets (Callithrix jacchus). J Neurosci 15,
7892-7904.
YOSHIMURA, Y. & CALLAWAY, E. M. (2005). Fine-scale specificity of cortical
networks depends on inhibitory cell type and connectivity. Nat Neurosci 8,
1552-1559.
YOSHIOKA, T., DOW, B. M. & VAUTIN, R. G. (1996). Neuronal mechanisms of
color categorization in areas V1, V2 and V4 of macaque monkey visual cortex.
Behav Brain Res 76, 51-70.
YOSHIOKA, T., LEVITT, J. B. & LUND, J. S. (1992). Intrinsic lattice connections of
macaque monkey visual cortical area V4. J Neurosci 12, 2785-2802.
YOSHIOKA, T., LEVITT, J. B. & LUND, J. S. (1993). Anatomical basis for channel
interactions in macaque cortical visual area V1. Invest. Opth. and Vis. Sci. 34.
YOUNG, T. (1802). On the theory of light and colors, London: The Society.
ZEKI, S. (1983a). Colour coding in the cerebral cortex: the reaction of cells in
monkey visual cortex to wavelengths and colours. Neuroscience. 9, 741-765.
144
CITATIONS
ZEKI, S. (1983b). Colour coding in the cerebral cortex: the responses of
wavelength-selective and colour-coded cells in monkey visual cortex to
changes in wavelength composition. Neuroscience. 9, 767-781.
ZEKI, S. (1983c). The distribution of wavelength and orientation selective cells
in different areas of monkey visual cortex. Proc R Soc Lond B Biol Sci. 217,
449-470.
ZEKI, S. (1983d). The relationship between wavelength and color studied in
single cells of monkey striate cortex. Prog Brain Res 58, 219-227.
ZEKI, S. (1990). A century of cerebral achromatopsia. Brain 113, 1721-1777.
ZEKI, S. (1991). Cerebral akinetopsia (visual motion blindness). A review.
Brain. 114, 811-824.
ZEKI, S. & MARINI, L. (1998). Three cortical stages of colour processing in the
human brain. Brain. 121, 1669-1685.
ZEKI, S., WATSON, J. D., LUECK, C. J., FRISTON, K. J., KENNARD, C. &
FRACKOWIAK, R. S. (1991). A direct demonstration of functional specialization
in human visual cortex. J Neurosci. 11, 641-649.
ZEKI, S. M. (1973). Colour coding in rhesus monkey prestriate cortex. Brain
Res 53, 422-427.
ZEKI, S. M. (1974). Functional organization of a visual area in the posterior
bank of the superior temporal sulcus of the rhesus monkey. J Physiol. 236,
549-573.
ZEKI, S. M. (1978a). The cortical projections of foveal striate cortex in the
rhesus monkey. J Physiol 277, 227-244.
ZHENG, D., LAMANTIA, A. S. & PURVES, D. (1991). Specialized vascularization
of the primate visual cortex. Journal of Neuroscience 11, 2622-2629.
ZIHL, J., VON CRAMON, D. & MAI, N. (1983). Selective disturbance of movement
vision after bilateral brain damage. Brain 106 (Pt 2), 313-340.
145
Curriculum Vitae
Curriculum Vitae
Personal Information:
Name
Birthday
Birthplace
Address
Family status
Nationality
Marcus Grüschow
08.09.1976
Magdeburg
Halberstädterstrasse 52, Magdeburg, 39112
Unwed
german
School Education:
1983-90
Hubert-Materlik Oberschule, Magdeburg, Germany
1990-93
Geschwister Scholl Gymnasium, Magdeburg, Germany
1993-94
Tolland Highschool, Connecticut, USA, Honorary
Diploma
Geschwister Scholl Gymnasium, Magdeburg, Germany
(Abitur)
1994-96
University Education:
Sept 1997- Sept 1998
Oct 1998- Oct 1999
Oct 1999- Sept 2003
Since Oct 2003
Otto-von-Guericke University, Magdeburg, Germany
(Major: Education, Minors: Psychology and English)
State University of New York at Oswego, USA (2
Semesters, Major: Psychology, Minors:
Communications, Percussions)
Otto-von-Guericke University, Magdeburg, Germany
(Major: Psychology); Dipl. Psych. Sept 2003
Ph.D. Thesis at the Faculty of Natural Sciences at the
Otto-von-Guericke-University Magdeburg and the
Center for Advanced Imaging
Magdeburg, February 2007
146